0:00

Hey, Big Biology listeners, it's fun

0:02

drive time. Please remember that

0:04

we're a nonprofit, and we need your help

0:06

to keep making the podcast you love. And

0:08

to stretch your contribution to its max,

0:10

a superfan of Big Biology is generously

0:13

providing up

0:13

to $10,000 in matching funds

0:15

for each donation. You get to double your

0:18

support of our Big Biology producers, interns,

0:20

and artists this time. So please

0:22

consider making a donation today. Otherwise,

0:24

the first 20 minutes of the next few shows

0:27

might sound like NPR does at this time of

0:29

year. So much air time on

0:31

fundraising. There are three ways to support

0:33

us. The first is a one-time donation, which

0:35

you can make at our website, bigbiology.org.

0:38

The second is to sign up as a patron at our

0:40

Patreon site, which is at patreon.com

0:43

slash big bio. And the last one is a free

0:45

one.

0:45

Just spread the word about the show. Twitter, Instagram,

0:48

whatever. If you like Big Biology, tell a

0:50

friend and let them pay our bills.

1:00

Light the fireworks and blow the trumpets.

1:03

It's the 100th episode of Big

1:05

Biology. That's a hecta episode for

1:07

those of you who tend to invoke Greek prefixes

1:09

unnecessarily. Okay, it's going to be that

1:11

kind of show, huh? Even episode 100? Have

1:14

you no respect, man? We considered

1:16

many options for our 100th episode, but in

1:18

the end, we decided that it would be fun to pick

1:20

out just one theme and piece together

1:22

how different guests have approached it.

1:24

We picked agency. A totally

1:26

non-controversial topic, right, Cam? Cam?

1:30

Cam? Yeah, we'll hear from Cam

1:32

later in the show. Let's just say he has different

1:34

opinions on agency than me and Marty. But

1:37

why did we choose agency? We chose it because

1:39

many past guests have convinced us that we really

1:41

should be paying it a lot more attention in our own

1:43

research. And all of that unintended cajoling

1:46

worked. We've come to think a lot more about

1:48

agency and its implications for the work we do. For

1:50

instance, it's really helped me to think about how epigenetic

1:53

marks like DNA methylation might enable

1:55

organisms to respond plastically and invade

1:57

new areas. And it's helped me design better

1:59

experiments. on niche construction and thermal

2:01

regulation in insects. Also agency,

2:04

or at least one of its core elements, information,

2:06

has so resonated with us that we've profoundly

2:09

changed the way we think about stress in animals.

2:11

Art and I have been interested in the topic for years,

2:13

and after talking about information and agency

2:16

to many guests on big biology, we've

2:18

changed our approach to this topic big time. But

2:20

in the broadest sense, we chose to focus on agency

2:23

in the 100th Big Biology episode, because

2:25

it illustrates why so many biologists

2:27

have called for updates to the modern synthesis.

2:29

Even extended evolutionary synthesis, some have called

2:32

it. They want a more complete theory of life. Unfortunately,

2:34

the message of the extended synthesis crowd

2:37

often came across as the original modern

2:39

synthesis was incomplete or wrong. Which

2:41

is not just off-putting, it misses the point.

2:44

The major innovation of the modern synthesis was to take

2:46

statistical shortcuts and thus make

2:48

tractable that which until then had

2:50

been out of reach. It made the complex

2:53

simple. Darwin's ideas about population

2:55

change over time were first merged mathematically

2:57

with the principles of inheritance that Mendel gleaned

2:59

from his pea plants. Then Fischer, Haldane,

3:02

Wright, and others invented concepts like

3:04

gene types, populations, and

3:06

relative fitness to predict genetic evolution

3:09

and voila, a lot of biology became

3:11

intelligible. And predictable. Based on

3:13

a few simplifying assumptions. Here's

3:15

Dennis Walsh, our guest from episode 62 explaining

3:18

the modern synthesis. The modern synthesis

3:20

is a really abstract

3:22

theory. You

3:25

take these abstract entities,

3:27

gene types, and you construct

3:29

an abstract entity, a population,

3:31

and an assemblage of gene types, and

3:34

then you apply to

3:36

the gene types this very abstract

3:39

para parameter, growth rate,

3:41

relative growth rate. You can track

3:43

the dynamics of these abstract populations

3:46

in this very, very powerful way. And

3:49

the- While skipping over the organisms. Exactly.

3:51

Exactly. Leaving them out. Right.

3:54

So, how does this parameter fitness or growth rate-

3:57

all those biological things-

4:00

packed into them. It can accommodate

4:02

any amount of biology. But

4:04

I think what the defenders of the modern synthesis

4:07

don't do is they don't acknowledge

4:10

or realize the level of abstraction

4:13

at which this theory actually works.

4:14

It's not about the

4:17

nuts and bolts biology. It's basically

4:20

identifying a universality

4:23

phenomenon that we find in

4:25

thermodynamics. Any time you have

4:28

an assemblage of differentially growing

4:31

or changing entities, you're going to have

4:33

this change

4:35

in the population of the ensemble

4:39

that's expressible at a really high

4:41

level of abstraction. And I think that's

4:44

the core of the modern synthesis.

4:45

That's what makes it so powerful. But

4:48

it's not a theory of

4:51

the metaphysics of evolution, or

4:54

as I say in the book, what happens when evolution

4:56

happens. It's a very poor account

4:58

of what happens when evolution happens.

5:01

In this light, it's strange to say that the

5:03

modern synthesis needs an update because it

5:05

always rested on what William James

5:07

called vicious abstraction. By

5:10

intentionally leaving out many of the details that

5:12

make life life, we learned

5:14

things we never knew and made all sorts of progress

5:16

in animal husbandry, improvements in crop

5:19

productivity, and even in human medicine.

5:21

That said, we empathize with calls

5:23

by some for an extended synthesis. We

5:25

agree that biology is more complex than the modern synthesizers

5:28

imagined. And we think it's time to strive

5:31

for a more inclusive and cohesive theory of

5:33

life, one that captures its origins, its

5:35

adaptive operations, and its

5:37

evolution. To channel Sarah Walker, guest

5:40

on episodes 9 and 93,

5:42

we need a model of life that focuses more

5:44

on the alive parts of life.

5:46

In my mind, life is a dynamical

5:48

process. And it's one where you have a

5:51

particular informational patterns that are like

5:53

structuring physical systems across space and time.

5:56

People get mad that I think a computer is life

5:58

or a screwdriver is life.

5:59

but those things literally would not be created without

6:02

information. Yeah, it was consistent. That's okay. And

6:05

then alive is the systems that

6:07

are actively constructing things. They're the ones doing

6:09

the information processing to actually build those

6:11

things and use internalized information

6:13

to actually do the construction. The modern synthesis

6:15

took shape at a time in history when inheritance

6:18

was just coming to be understood, and the

6:20

molecular revolution, including the discovery

6:22

of DNA, was just getting underway. Progress

6:24

via synthesis thinking was great, but it

6:26

also led many subfields of biology to forget about

6:28

the original vicious abstraction. Most

6:31

egregiously, genes came to take on

6:33

causal powers they simply can't have,

6:35

including the culmination of this forgetfulness,

6:38

the selfish gene concept. To Richard

6:40

Dawkins, George Williams, Arvid Oggren, a

6:42

guest on episode 73, and many

6:44

others, willow

6:45

catkins blowing from trees are literally

6:48

DNA rain. To their way of thinking, one

6:50

need only understand how the DNA from

6:52

one willow generation gets into the next.

6:55

That's it. That's all. Biology done.

6:57

Nuh-uh.

6:58

Can't be. The modern synthesis can't

7:00

be the main paradigm for understanding

7:02

life if it doesn't explain most of how

7:05

life is alive, how complex systems

7:07

emerge and come to be resilient, how homeostasis

7:10

works, and how collections of sub-entities

7:12

are integrated into working holes. One

7:14

of our very first guests, Massimo Piliucci

7:16

from episode 7, beautifully laid out

7:18

the limits of the modern synthesis. But

7:21

to be fair, he was using Dick Lewington's analogy.

7:23

So he said, look, imagine you're

7:26

building a house,

7:27

and instead of in the United States where most

7:29

houses are built of wood, which is why they

7:32

don't last, you

7:34

build it the old-fashioned way, the European

7:36

way, with bricks and lime. So

7:39

you say, okay, so you start putting

7:41

the

7:42

first layer of bricks and

7:44

then lime on it, and then bricks and lime

7:46

and bricks and lime. Now, once you get the final

7:48

house,

7:50

you could, if you wanted, ask the

7:52

very quantitative question, well, what

7:55

is the weight of the house in bricks?

7:57

And what's the weight of the house in lime? And

8:00

there is an answer to that question. And I'm sure it will be

8:02

something like 98% bricks and 2% line. That

8:07

tells you precisely nothing about how to build

8:09

a house. Because it isn't

8:11

about, you're not going to come up with 98 bricks

8:14

and then two little pieces of line

8:17

and then you say, oh, I got the house. No, you

8:19

get the house by the specific patterning

8:22

of the bricks and the line. And so the

8:24

idea there is that even if you could

8:26

show that, let's say 90% of

8:28

variation in phenotypes

8:31

in a particular human trait is the result of genetic

8:34

influences,

8:35

that still doesn't mean the way

8:38

in which is usually interpreted. Oh, so

8:40

genes do all the work and the environment is not important.

8:43

You take out that 10% in that specific

8:45

patterning, you get nothing, absolutely nothing.

8:48

Because genes by themselves don't do crap. And

8:50

yes, you can buy that on a bumper sticker

8:53

in the big biology store. Massimo was

8:55

talking here about the importance of phenotypic plasticity,

8:58

but his point applies more broadly. The

9:00

modern synthesis focuses on changes in the

9:02

fraction of variation that things, either

9:04

genes or environments, explain. But that

9:06

kind of model doesn't capture much of what we wanted

9:09

to understand in the first place. How

9:11

the house comes to take the form it does, and

9:13

how the house doesn't fall apart soon after

9:15

being built.

9:16

Can we come up with another theory of life? A

9:18

simple one, but also one that better captures

9:20

the key things that distinguish life from non-life.

9:23

We think we can. And we think at its core

9:25

will be the concept of agency.

9:27

A system's propensity to maintain its integrity

9:30

by either changing the disruptive external forces

9:33

it experiences,

9:34

or adjusting its internal makeup to better

9:36

suit those challenges and opportunities. Or

9:39

to blend the words and ideas of a bunch of past

9:41

guests, agency is the intentionally

9:43

cognitive set of activities that a living

9:45

system uses to achieve dynamic

9:47

stability. The architecture and updates

9:49

of the Bayesian Priors and its Markov blanket

9:52

model of the world.

9:53

In this 100th episode of Big Biology, we focus

9:55

on agency with input from a subset

9:57

of our past guests to try to forecast where

9:59

biology might be.

9:59

be going and what values lie

10:02

in the conceptual transformation. And in the

10:04

last third of the show, Art and I will talk with the newest

10:06

host of Big Biology, Cam Gallimbor, on

10:09

his thoughts and skepticism about agency. I'm

10:11

Art Woods, and I'm Marty Martin, and this

10:13

is Big Biology Episode 100!

10:29

To make a case for agency, let's start at the literal

10:32

beginning, life's origins. In Episode 49,

10:34

we talk with Nick Lane about this topic and his

10:36

book, The Vital Question.

10:38

Most origins of life scientists, at least historically,

10:41

identify with one of two camps, the

10:43

RNA first world or the metabolism first

10:45

world.

10:46

Nick has feed in both worlds with heavy

10:48

doses of systems and thermodynamics

10:50

thinking. The old idea of a primordial

10:53

soup, it's not gone away.

10:55

Darwin's warm pond is probably still

10:57

the dominant idea in the origin of life

11:00

field, except when we would now call it terrestrial geothermal

11:02

systems. And

11:05

a lot of that chemistry works quite well.

11:08

You start with cyanides or cyanosetylene,

11:10

you use UV radiation, and you're able to make

11:12

all the building blocks of life. And

11:15

the problem for me with all of that

11:17

is

11:18

beautiful chemistry that works well, but

11:20

it doesn't look anything like biochemistry. And

11:22

so you're still left with this question, okay, so we've got all

11:25

these all these monomers floating around in

11:27

solution.

11:28

What happens next? How do they invent life

11:30

from there? As you can hear, Nick wonders

11:32

whether particular conditions on young Earth led

11:35

to the appearance of simple but self-sustaining

11:37

complex systems. Data from his and

11:39

other labs show that proto-life processes

11:41

were originally possible because of the availability

11:44

of particular forms of matter and energy in

11:46

particular places. Many sites on early

11:48

Earth would not have supported lifelike systems, but

11:51

some might have been just rich in resource dense

11:53

enough for those systems to maintain their integrity

11:55

and eventually reproduce.

11:57

One way, perhaps the way, that these systems

11:59

were able to...

11:59

maintain integrity was by instantiating

12:02

information into their structural elements,

12:04

which include the nucleic acids that became

12:07

today's RNA and DNA. Now it remains

12:09

a mystery how lifelike systems came alive,

12:12

from this point of being just a dynamically stable

12:14

system to the complex homeostatic modulus

12:17

systems that resemble modern microbes. Probably

12:19

there was no bright line between these stages anyway.

12:22

An early step in this process must have entailed

12:24

a separation of internal and external states,

12:27

some sort of physical boundary like what would become

12:29

a cell

12:29

membrane. Exactly what that first membrane

12:32

would have been, much less how the inside

12:34

came to take on such diversity, is

12:36

to put it lightly, a work still in progress.

12:39

Nevertheless, according to Carl Friston, our

12:41

guest in episode 70, all viable

12:43

complex systems have what mathematicians call

12:45

Markov blankets. The first barrier

12:48

must have provided a clear demarcation between outside

12:50

and in. And to

12:51

be clear, Carl isn't referring to Markov's

12:53

blanket as the thing he slept under during particularly

12:56

cold Russian winters. Rather Markov blankets

12:58

are relatively simple interconnected sets of states

13:00

whereby complex systems shield themselves

13:03

from entropy. At its simplest, a Markov

13:05

blanket is just a way

13:08

of partitioning the states

13:11

of a universe into

13:14

a system of interest, say

13:17

you or me or the virus or the vegan

13:19

and everything else.

13:21

More specifically, it's

13:23

a partition, a dividing

13:26

into three sets

13:29

of states, those states that are internal

13:31

to a system,

13:33

those states that are external

13:35

to the system, and then some intervening

13:37

states that mediate the exchange

13:40

between the inside and the outside.

13:42

So if you were a physicist

13:45

or if you go back to your schoolboy

13:47

physics, the Markov blanket enshrouds

13:50

the internal states. If you're a biologist,

13:53

then you can think of this as the cell surface.

13:55

It's the thing that sort of together with the internal

13:58

states constitutes the unit

14:00

of description of discussion

14:04

and is responsible for mediating

14:07

the reciprocal, the two-way

14:10

causal exchange between the inside and

14:12

the outside. Mathematically, it

14:14

inherits from the work of Pearl

14:18

in Bayesian networks and is defined

14:21

operationally in terms of what's called conditional

14:23

independence, which means that quite

14:25

simply,

14:26

if I wanted to know how my

14:29

internal states are going to change

14:32

in the future, in

14:34

the immediate future, then I only

14:37

need to know

14:38

the Markov blanket states, the surrounding

14:41

states. I don't need to know the rest of the universe.

14:43

Carl says that Markov blankets are the key elements

14:46

in his free energy principle. The idea

14:48

that all enduring complex systems, including

14:50

living ones, must resist entropy

14:52

by minimizing surprise.

14:55

Carl's surprise here is mathematically

14:57

defined. He means how unexpected

14:59

information gleaned from the environment, internal

15:01

or external, is related to a set of Bayesian

15:04

priors. These are ineffective parameters that

15:06

comprise the Markov blanket. Here's Carl

15:08

on surprise

15:08

and free energy minimization, ideas

15:11

he attributed to Richard Feynman. So he was dealing

15:13

with a problem of trying to characterize

15:16

the behavior of small

15:18

particles in quantum electrodynamics, trying

15:22

to understand the probability distributions

15:24

or the beliefs about different paths

15:26

that particles could take,

15:29

realize and describe it properly.

15:32

He had to turn what was an impossible

15:34

integration problem, marginalization

15:37

problem, into an optimization

15:39

problem that he could then solve using

15:41

standard techniques. That's a key

15:43

move. What that does is it

15:46

creates, it takes a

15:49

system that can be described probabilistically,

15:51

in this instance, quantum mechanics,

15:54

and converts it into an

15:56

object that can be understood

15:59

in terms of

15:59

of optimization. And that

16:02

means you've now got a normative, theological

16:06

gloss on describing how this

16:08

system works, because it looks as if something

16:10

is being minimized or maximized.

16:13

So what is that thing? Well, the thing

16:16

is the variation for energy. So it

16:19

is exactly the same construct

16:21

that is using machine learning and high

16:23

end deep learning like variational autoencoders,

16:26

where the negative

16:29

physics Feynman variational free energy

16:31

is known as an evidence lower bound and an elbow.

16:35

So in

16:36

that name, in that acronym,

16:38

you have the key thing, which

16:41

is evidence. So what

16:43

we are talking about now is a

16:46

generic mathematical way of

16:48

describing the

16:50

probabilistic dynamics or evolution

16:52

of any system in a normative

16:55

sense, in a theological sense, as trying

16:58

to optimize

17:01

a bound on evidence. So what

17:03

is evidence? It's just the probability

17:06

of some outcomes,

17:08

given some model

17:10

or hypothesis or construct

17:13

explanation that

17:15

you consider generated

17:18

those outcomes. If Carl's

17:20

hypothesis that all complex systems have

17:22

Markov blankets, all the way down,

17:24

as he likes to say, lifelike systems have

17:26

to make choices to endure. They must

17:29

either push back on the world to make it what they need it

17:31

to be, or they must update their models

17:33

of the world. If their models become sophisticated enough

17:35

that they can come to plan about how to push back,

17:38

they

17:38

by definition have agency.

17:40

Given enough time and the right contexts, one

17:43

should expect such systems to become common.

17:46

One more time. Here's Carl. If you

17:48

believe that the universe is

17:50

essentially a random dynamical

17:52

system, what you are saying is that the

17:56

variables and states of the universe evolve,

17:58

which means that a trajectory.

18:01

If they have a trajectory, then

18:04

if you want to understand that

18:06

trajectory in a normative sense, it's

18:08

basically doing some form of gradient descent,

18:11

some sort of either hill climbing or hill descending,

18:13

which gives you this sort of teleological optimization

18:16

perspective. But in so doing, even

18:19

an elementary particle

18:22

is in effect selecting

18:24

a path to pursue.

18:26

So in that elemental sense, there has

18:28

to be agency because there's time, you

18:30

can't have dynamics without time. And

18:32

if there's time, then there are trajectories.

18:35

If there are trajectories, then I go over there

18:37

and I don't go over there. So I

18:39

agree entirely that you can't

18:42

move away from or deny an

18:45

agentual aspect even to elementary

18:47

particles. Having said that, I

18:50

think there is a fundamental difference between

18:52

agency that involves planning and

18:55

agency that is just an expression of

18:57

density dynamics. So my favorite

19:00

example is the difference between a virus and a vegan.

19:02

The virus certainly

19:05

has attained

19:07

a non-equilibrium steady state. It's a beautiful

19:09

little model of its eco-nish.

19:13

It's milieu in which it survives. It does

19:15

all the right things entailed

19:17

in its sort of

19:20

molecular structure and kinetics

19:22

are all the right

19:25

substrates to be interpreted

19:27

as a model of the kind of inputs and outputs

19:30

in the world, the intracellular world usually

19:32

that it inhabits. And

19:35

one could say the same as a vegan, but a vegan

19:38

of course can do a lot more than a

19:40

virus. And if you can plan, then

19:43

by definition you must have a generative

19:45

model

19:46

of the consequences of your action in

19:48

the future. And of course in the

19:51

future now says well

19:53

you've got a generative model with the temporal depth

19:55

with the horizon. So then you

19:57

can ask well how far into the future can you go?

19:59

I see? Well,

20:01

in a sense, a virus could probably see a few

20:03

nanoseconds or milliseconds into the future simply

20:05

by committing to a particular trajectory. But

20:08

of course, a vegan

20:09

can not only see a few milliseconds

20:11

into the future, she can see

20:15

minutes, hours, days, months in

20:17

terms of- Half a century, yeah. Half

20:19

a century, exactly.

20:26

Okay, so combining Nix and Carl's thinking

20:28

so far, the first forms of life on

20:30

Earth must have been self-sustaining systems. Or

20:33

autocatagletic sets, in the words of Stuart

20:35

Kaufman. Appearing in one or more places conducive

20:38

to persisting.

20:39

To avoid succumbing to entropy, the Markov

20:41

blankets of successful systems would have become

20:43

more and more complex. If they couldn't

20:45

change their external states, they'd

20:47

instead update the parameters of their internal

20:50

states, eventually instantiating information

20:52

about past successes into nucleic acids

20:55

and probably many other structures. Paul

20:57

Davies, guest on episode 33, and

20:59

author of many books, including the focus

21:02

of our chat, Demon and the Machine, said that this

21:04

use of information, in the sense Carl talked about,

21:06

is one of the things that distinguishes life

21:08

from non-life. I think everybody

21:11

thinks about the world about them will realize

21:13

that life stands out. Living things

21:16

are in a class apart. They perform

21:18

the most amazing feats.

21:20

They seem to be different in

21:22

a very fundamental way, and not just difference

21:25

in degree, but difference in principle

21:27

from non-living systems. They are

21:30

very, very odd. Now, biologists

21:32

tend to not

21:34

find their subject matter quite so

21:36

odd, because of course they're dealing with it every day. Life's

21:38

what they study, so they sort of take it for granted. Of course, of course

21:40

we know what it is. To a physicist, it looks

21:42

like magic. It really does.

21:44

I remember thinking, when I

21:47

suppose, if you take a living organism,

21:50

it's made up of normal atoms

21:52

doing normal physics things.

21:55

How is it that a collection of

21:57

these stupid atoms, blundering

21:59

around just for

21:59

following the rules of physics, can collectively

22:03

combine to produce some,

22:05

what looks like magic, some form of magic. How

22:08

can that happen? And it is a very,

22:10

very profound mystery. To Paul, whereas

22:13

we're taught from our first biology classes that

22:15

DNA encodes

22:16

information, what that really means

22:18

is incredibly important for understanding life.

22:21

Without doubt, some kind of information is in DNA,

22:24

but information is everywhere in life.

22:26

What's in DNA is not all of it. And even

22:28

in DNA, the information that resides in

22:30

the sequence isn't necessarily even the most

22:33

important bit. Here's Paul again. You're

22:35

absolutely right that the information

22:38

contained in DNA, the genetic information,

22:41

is something that people are familiar with,

22:44

but it doesn't

22:46

stop there. So genes rarely act

22:48

in isolation.

22:49

They can switch each other on and off,

22:51

and they can form networks sometimes of great

22:54

complexity, and information swirls

22:56

around these networks. And sometimes

22:59

they're very much like components

23:01

in some electronic system.

23:04

They form modules, and these modules,

23:06

in turn, couple to each other, form

23:08

bigger networks. And we're

23:11

talking here, unlike in electronics,

23:13

about components

23:15

being wired together chemically, not electrically.

23:18

But the same principles apply, that

23:20

these are logical operations that

23:23

these components can carry out, and they can

23:26

compute and regulate and

23:29

fulfill many of the functions of modern electronics

23:32

and computing. But they're doing it with

23:36

a chemical basis. And so genes

23:38

form networks, but it

23:41

doesn't stop there because cells themselves

23:44

can form communities. They can signal each

23:46

other chemically. So we're using this information

23:49

language all the time. We talk about

23:51

cell-cell signaling cooperation

23:53

among colonies of cells. Even bacteria

23:57

can form communities that

23:59

can carry out coherent tasks.

24:03

And then when we come up to multicellular

24:06

organisms, take social insects for

24:08

example, one of the really fascinating

24:10

areas of study here at Arizona State University

24:13

is with ants and ant communication.

24:15

And they form colonies and they engage

24:18

in collective decision making.

24:20

You can see these pictures where ants

24:22

are sort of clustering around, you know,

24:24

having a little conference and you're

24:26

wondering, you know, what are they talking about. And

24:30

we're beginning to understand now, there's

24:33

no sort of chief ant that

24:35

says, you know, okay

24:38

lads and it's not lads because they're all female gals,

24:43

you know, we're off to a new nest. It's one of these things

24:45

that is done collectively, it's distributed

24:47

across them and it's done through information

24:50

exchange through all sorts of chemical

24:52

and physical cues. And it goes on all

24:54

the way up. We've talked about the brain, this

24:57

is the biggest information processing system

24:59

that we know. But again,

25:01

it doesn't stop there. It really encompasses the

25:04

entire planet.

25:05

When we look at ecosystems,

25:07

there's a lot of information flow. There

25:09

are mobile genetic elements, things like

25:12

viruses that get around the environment, couple

25:15

widely separated systems together. So

25:17

I like to say that the biosphere

25:19

is the original world wide web. The problem with

25:21

adding information to our theory of life is that

25:24

it's a very abstract concept. So

25:26

we've had a hard time measuring it. In the 1950s, for

25:29

instance, cybernetics was all the rage

25:31

and many scientists argued that by studying information,

25:34

the mysteries of life

25:35

would soon be solved. Wrong. Although

25:37

in some context, information theory was really fruitful.

25:40

We're looking at you neuroscience. In most

25:42

biological sub-disciplines, it just never delivered.

25:45

Information was truly a bad idea or was too

25:48

diffusely defined, leading people to talk past

25:50

each other, or it was just never really measured

25:52

well in any of its possible forms. Several

25:55

guests now feel that the tides have turned and

25:57

that information will in fact be integral

25:59

to biology's future.

26:00

For instance, Carl Friston thinks that the information

26:03

geometry of a system will be what distinguishes

26:05

a stone from a virus.

26:07

Sarah Walker, on her second visit in

26:10

episode 93, seemed to agree. She

26:12

and collaborator Lee Cronin proposed that something

26:14

called the Assembly Index will be useful

26:16

to finding extraterrestrial life. These

26:18

indices can be calculated in principle for

26:21

anything, and inherently they represent

26:23

the kind of thing Carl meant about information

26:25

geometry. Sarah used Harry Potter's Hogwarts

26:28

Castle, one made of Lego, that is, as

26:30

an example.

26:31

Sarah thinks that the higher the Assembly Index

26:33

is of a thing, the more likely it was produced

26:36

by non-random processes, namely life. Only

26:39

information rich things like life could

26:41

build high Assembly Index structures.

26:43

So, with Hogwarts Castle, for example,

26:46

imagine I had just put the Legos on the table, and

26:48

I didn't give you the instructions. And maybe

26:50

you're a child that never read Harry Potter,

26:53

and I said, make Hogwarts. What is

26:55

the likelihood of you even being

26:57

able to build that object? So, the

26:59

fact that you could even imagine the experiment probably

27:01

means that you have some cultural

27:03

association with my cultural background. But

27:06

there was a goal in mind, and you can imagine building

27:08

toward that goal, and probably you were assuming you had the

27:11

Lego instructions in front of you to build it. So,

27:13

Hogwarts has a very high Assembly Index. The

27:16

minimal path to make Hogwarts by randomly

27:19

constructing it, just by joining operations, is

27:21

quite large. You know, if I had said,

27:23

let's just stick three blocks together, red, blue,

27:26

red, you know, that would be pretty easy

27:28

for you to randomly assemble. And

27:30

so, the idea is that everything can be tiered by

27:32

this minimal path, which we call

27:33

the Assembly Index, and the things that have

27:35

a larger depth in time, require

27:37

more minimal steps, more memory to produce them, are more

27:40

evolved objects. They require more evolution

27:42

to get to them, more knowledge, more learning.

27:44

If you heard that episode, you could probably

27:46

tell that I was skeptical about Sarah's Assembly

27:48

Index as Panacea. Or I'm too dense to understand

27:50

her. Yeah, she's way smarter than you. abstraction

28:01

extinguished the fire of life. It's dynamism.

28:03

I understood that even life processes like metabolism

28:06

could be captured by the assembly index,

28:08

but it wasn't convinced that an idea developed to identify

28:11

alien life was sufficient to explain

28:13

living life. I

28:14

mentioned this to Sarah and she said assembly theory

28:16

accounted for it, but I just couldn't quite accept her explanation.

28:19

Well, for once you're in good company, Marty, as

28:22

this very thing is what motivated Dan Nicholson,

28:24

guest on episode 82, to co-edit

28:26

the book Everything Flows. To Dan,

28:28

a key trait of life is its dynamic equilibrium.

28:31

It's for this reason that we titled his episode,

28:34

Organisms are Not Machines. In

28:36

other words, life is more a river than

28:38

a riverboat.

28:39

Organisms are, from a physical perspective, these

28:42

systems, okay, they're systems that have

28:45

to maintain their organization by

28:47

constantly bringing in matter and energy from the environment

28:49

so that they attain this steady state. If they lose

28:51

the steady state, they die. It's an irreversible

28:53

process and that you can think of everything

28:56

that organisms does ultimately

28:58

is being reducible to that, even if that in practice may

29:00

not necessarily be helpful, but you know at

29:02

least that whatever else organisms

29:05

are, what can't be denied is that they're self-organizing

29:08

systems. I mean, it's puzzling to me because it

29:10

isn't something that often comes

29:12

up in biology. If you really press a

29:14

biologist, that's like, well, of course, and yet it

29:16

doesn't usually feature in the way biologists

29:19

explain certain phenomena. It's a reminder

29:22

that you can provide a physical

29:25

explanation for certain

29:26

capacities that organisms have in

29:28

a way that shouldn't be controversial, shouldn't

29:30

be problematic, and yet it's a way of thinking

29:33

about them physically that is very different from the

29:35

traditional mechanistic, reductionistic, deterministic

29:38

view that has dominated the biological

29:40

discourse since the 17th century. So it's saying,

29:42

okay, I'm giving you an alternative and

29:44

I'm going to anticipate your

29:48

protest that is not scientific by saying I'm grounding

29:50

it in physics and I'm going to show

29:52

to you that

29:53

this grounding leads to really interesting

29:56

implications for how you should think about biology.

29:58

Then push so hard against like... as things,

30:01

namely machines, because this metaphor

30:03

has so biased the field. Machines

30:05

like cars don't get tired, they don't rebuild

30:07

their own tires when the old ones wear down, and they

30:10

don't make baby cars. Organisms

30:12

do all of these things, and when we abstract organisms

30:15

into simpler systems like machines

30:17

or worse, genetic blueprints, we

30:19

miss the main points about life. We

30:21

want a simple model of life, because a too

30:24

complex model defeats the purpose. Okay,

30:26

first, before we dive into agency, we

30:28

have to say that, yeah,

30:29

we know, we get it, that agency

30:32

is a taboo topic to many biologists. It

30:34

smacks of some god of the gaps, some

30:37

form of soul, or a spooky force

30:39

galvanizing ourselves.

30:41

Historically, though,

30:41

agency and other vitalistic ideas

30:44

didn't leave the bad taste that they now do. Claude

30:47

Bernard, Louis Pasteur, and many

30:49

others had much more sophisticated and nuanced

30:52

ideas of agency than the simplistic negative

30:54

one that's

30:55

so common now. Hopefully by this point in the episode,

30:57

you can tell that the agency we mean is not

30:59

a mystical spiritual one. We see no

31:01

room for such things in life, and we're both

31:04

fairly staunch atheists. However, we and

31:06

the many guests we quoted earlier do think

31:08

that agency, or something like it, promises

31:11

the chance to truly and fully integrate

31:13

biology. Maybe not all the details,

31:16

but all the major elements. So

31:18

let's get after it with help from past guests,

31:21

starting with Dennis Walsh. Dennis' book,

31:23

Organisms, Agency, and Evolution, had a huge

31:25

effect on us. To Dennis, for life to

31:27

sustain itself and evolution to occur,

31:30

agents must seek out resources, avoid

31:32

danger, and generally expose themselves

31:34

or not to selective pressures. Some things

31:37

are inherited. Memory tokens, as Scott

31:39

Turner called them, and these factors plus

31:41

DNA affect change over generations.

31:44

But what happens within generations, organisms

31:46

struggle for existence, is what Dennis thinks

31:49

needs a lot more attention than it gets. In

31:51

other words, if organisms didn't have agency,

31:53

they'd never succeed in this struggle. Organisms

31:56

as systems, always at risk

31:58

of breaking down, must have agency. to

32:00

explain what Dennis calls affordances.

32:03

So I think the affordance concept is really

32:05

important. It helps us

32:07

to explain the external

32:10

dynamics of organisms, how they move through their

32:13

environments and why. But

32:15

they're internal dynamics too. Why the parts

32:18

are integrated in the way they are. Why

32:21

organisms synthesize these very

32:23

materials out of which they're made. Because

32:25

they're conducive to the pursuit

32:28

of the organisms' goals and the exploitation

32:30

of their affordance. And they also create

32:32

affordances. The

32:35

structures or capacities confer

32:38

on

32:39

organisms capacities to

32:41

capabilities to pursue their

32:43

lives in this particular way. So

32:45

we should understand the integration of organisms

32:48

and their movement through their environment in terms

32:50

of the creation and exploitation

32:53

of affordances. And affordances

32:55

are dynamic. As you respond to an affordance,

32:58

other affordances open up.

33:00

So there's this constant

33:02

creation and exploitation of affordances

33:04

going along. So my thought

33:06

about agency was we should start there. See that

33:09

knowledge that this is what organisms are like. This is

33:11

the kind of defining feature of life. And

33:14

see how working from the

33:16

taking the affordance notion is basic.

33:20

Transforms our understanding of the dynamics

33:22

of evolution. One of our other repeat

33:25

guests on Big Biology, episodes 39 and 65, is

33:27

Mike Levin. In

33:29

our first chat with Mike, we talked about the inheritance

33:31

of body form in multicellular organisms.

33:34

Convention, of course, has it that a developmental program

33:36

somehow resides in the genes, one

33:38

that unfolds over some period of time to

33:40

produce a mature body plan.

33:42

But not always true. At least not in flatworms

33:44

and frogs. Inheritance of their body plans

33:47

has to do with, believe it or not, electrical

33:49

fields. But that's for you to go here in episode 39

33:51

if you like. Our chat with Mike on episode 65

33:54

focused on agency, or more accurately,

33:57

something he called cognition and agendas.

33:59

and was largely based on a 2020 article

34:02

that he co-authored with Dan Dennett in Eon

34:04

magazine, called Cognition All

34:06

the Way Down. The main point of the article was

34:08

that only agents can have agendas.

34:11

Put a mouse and a ball on the top of a very steep

34:13

hill, and both might initially resist rolling

34:15

down the slopes. But shake that hill enough, and

34:17

only one system stays in place. Or

34:19

it climbs back to the top. Mike and Dan realize

34:22

that any system with an agenda also

34:24

requires a form of cognition, some

34:26

way to identify options and choose among

34:28

them to remain in the same state or to

34:30

exploit another one. This choosing of options is

34:32

what Mike means when he says cognition. And

34:35

cognition is core to what agents do. Here's

34:37

Mike talking about a molecular reaction, arguing

34:40

that in some sense even molecules have

34:42

simple forms of cognition. People

34:44

sometimes say to me, especially, let's say molecular

34:46

biologists, will sometimes say, look, you're

34:49

talking as if this thing made decisions, whatever.

34:52

But that's just the metaphor, right? You don't really mean

34:54

that. You know, I make decisions. This thing is just chemistry.

34:57

And I think it's very important. I think

34:59

that's a major mistake that you and I think

35:01

it's really important to get

35:04

from the get-go that I think agency

35:07

and cognition are a continuum or

35:09

a spectrum. To Mike,

35:10

cognition can and should be broadly defined.

35:13

The most conspicuous form involves the brain,

35:15

but really any system that has an information

35:17

geometry can be understood as cognitive.

35:20

So subsequently, even simple systems like

35:22

molecules and rocks can be cognitive in

35:24

that their interactions with other entities can lead

35:26

to predictable changes in their form. For me, agency

35:29

is a kind of center of gravity for

35:32

things like decisions, preferences, memories,

35:35

and in the more advanced implementation, sometimes things

35:37

like blame and credit and other

35:40

things like that.

35:41

Agents

35:44

are things that can make mistakes. Chemistry

35:46

and physics doesn't make mistakes. It just sort of does what

35:48

it does. But agents are, and I

35:50

think this is a point that Dan has made before, that

35:52

agents are capable of making mistakes.

35:55

And so there is a

35:57

kind of a whole spectrum of

35:59

different. different levels of sophistication

36:01

that different agents can achieve.

36:04

So the thing with

36:07

the ball and the mouse on top of a hill is

36:09

just an example of this. If you've got

36:11

a ball at the top of a hill, you can use

36:14

equations that will tell you what it's going to do. And

36:16

those equations have almost no

36:19

reference whatsoever to information processing, to

36:21

memory, to learning, to preferences. You don't need

36:23

any of that. You have a much simpler model that does

36:25

pretty much everything you want to do to predict

36:27

what that system is going to do. If you've got a mouse

36:30

at the top of a hill, Newton's equations

36:32

about what it's going to do with it, where it rolled

36:34

down the gravity well, unless the mouse is dead,

36:37

are almost useless because

36:39

now if you really want to understand what that system is

36:41

going to do or modify it and

36:43

make the mouse go somewhere else, you have

36:46

no hope other than through a

36:48

model that takes seriously what

36:50

that system actually is. And it's a system with preferences,

36:53

with memories, with all kinds of internal states

36:55

that are going to determine what happens later

36:57

on. Okay, so now that we've laid out what agency

37:00

is, let's confront the elephant in the room.

37:02

Does it matter? Do

37:04

we have any evidence that agential thinking will

37:06

contribute new and important things to biology?

37:09

Because we started the show today by highlighting the vicious abstraction

37:11

at the heart of the modern synthesis, let's start

37:13

there with the evolutionary implications

37:15

of agency. Specifically, let's flesh out the

37:17

roles that agency plays in what's become a hot

37:20

topic in evolutionary physiology. Thermal

37:23

regulation. Boo, how do your boys

37:25

put up with those terrible jokes? They love

37:27

me. Ectotherms often

37:29

use behavior to get body temperatures they want.

37:32

And recent thinking on this topic is

37:34

focused on something called the Bogart Effect,

37:36

which is named after the American herpetologist

37:38

Charles Bogart.

37:39

In the 1940s and 50s, Bogart studied

37:42

lizard thermoregulation and found something

37:44

surprising. Lizards in many different geographic

37:46

localities, living under many different prevailing

37:48

thermal regimes, had body temperatures that

37:50

varied little, primarily because

37:53

many of those lizards were such good behavioral

37:55

thermoregulators. Bogart wondered whether

37:57

behavioral thermoregulation could blunt the effect of the human's behavior.

37:59

of selection on other aspects of their thermal

38:02

biology, specifically their upper thermal

38:04

tolerances. And indeed, it can.

38:07

Here's Martha Munoz in episode 81 explaining

38:09

her work on the Bogart Effect.

38:10

Ray Huey, in collaboration with Paul

38:12

Hertz and Barry Senervo, wrote

38:15

a really impactful, conceptual

38:18

paper

38:19

taking Bogart's qualitative ideas

38:21

and creating a quantitative hypothesis

38:24

testing framework with which to put them to

38:26

the test.

38:27

If we take Bogart's argument

38:29

at the broadest possible level, what the

38:31

argument says is that when

38:34

any kind of regulatory homeostatic

38:36

behavior is at play, that has the capacity

38:39

to reduce environmental variation across

38:42

some environmental gradients, and that that

38:44

buffering should limit physiological

38:47

divergence and or slow

38:49

the rate of evolution.

38:51

Ray Huey and so Ray

38:53

and colleagues effectively

38:55

just gave this a new name,

38:57

the Bogart Effect, and devised a series

39:00

of approaches for testing it. And

39:03

what they did was basically

39:04

develop two premises that should be

39:06

true under the Bogart Effect. The first is that

39:09

regulatory behavior is occurring.

39:11

That, while it seems quite obvious,

39:14

is very complicated to actually demonstrate

39:16

in the field. It requires understanding

39:18

the

39:19

environment that's available to organisms.

39:21

So you need a null distribution of temperatures

39:23

that organisms, if you're doing thermal regulation, that

39:26

organisms could theoretically play with. Then you

39:28

need to demonstrate that

39:30

body temperatures are actually

39:32

in a mean and range that

39:34

is so separate from what's available

39:37

in the environment that our metrics indicate

39:39

that they're regulating. And so the null distribution

39:41

tells you what the environment is, and then you can

39:44

compare that to observed body temperatures

39:46

and through a series of metrics

39:49

basically quantify the degree

39:51

of thermal regulation that organisms are engaging

39:54

in. And this is step one. So

39:56

the second premise that should be true under

39:58

the Bogart Effect is that thermal regulation should

40:00

be associated with limited physiological

40:02

divergence or weaker selection on

40:04

physiology. The short version of Martha's

40:06

work to put a fine point on it? I've

40:09

discovered that lizards from tropical

40:11

islands tend to thermal regulate more than

40:14

lizards from the Latin American mainland.

40:16

And we discovered that the rate of heat tolerance

40:19

evolution is about three and a half times

40:21

slower on islands than on the

40:23

mainland.

40:24

Get that? 3.5 times slower evolution

40:26

on islands because island lizards do

40:28

so much more thermoregia regulation.

40:30

Their physiology is shielded from selection.

40:33

Although Martha didn't explicitly use the word agency

40:36

when she talked to us, we think it fits well into

40:38

the arguments we're making in the episode. To recast

40:40

Martha's work explicitly in this language, we

40:42

would say that lizards are agents, moving

40:45

around in their environments and exploiting affordances.

40:48

By which I mean choosing microclimates that give

40:50

high body temperatures when it's otherwise cool, and

40:53

microclimates that give low body temperature when

40:55

it's otherwise hot.

40:57

The outcome of this agency is remarkable

40:59

differences in the macroevolutionary trajectories

41:01

of lizard physiology in different lineages.

41:04

Martha's not nearly the only biologist that feels this

41:06

way. And perhaps not surprisingly, the biggest

41:08

proponents of agency as an important biological

41:10

force tend to be the physiologist, especially

41:13

those focused on homeostasis, how a system

41:15

maintains stability. Think thermostats.

41:18

Scott Turner from episode 36 said

41:20

that adaptations generally don't

41:22

really make sense except in the light of agency.

41:24

What we so liked about Scott's approach is

41:26

that he weaves

41:27

together adaptation and physics

41:29

with homeostasis as his threat. To Scott

41:32

and to Claude Bernard, a contemporary of Darwin

41:34

and the father of modern experimental medicine, homeostasis

41:37

is the process that distinguishes life

41:39

from non-life.

41:40

The very persistence of an organism's form

41:42

is itself a form of homeostasis.

41:45

And that, of course, is maintained by

41:47

this enormous complex of adaptive barriers

41:50

that separates us from the environment, the

41:53

lightings of lung, the lightings of the intestine,

41:56

the sensory interfaces, and

41:59

those kinds of things. all of which are mediated

42:01

by epithelium-like

42:04

structures. And you can take some

42:06

fairly simple aspects of

42:12

conservation of mass and thermodynamics to

42:15

be able to extend

42:19

adaptive boundaries outward from

42:21

the organisms. And in the

42:23

case of the termites, of course, these

42:26

are the African termites that build these

42:28

massive mounds as

42:30

infrastructure for their sub-training colonies.

42:34

What these mounds are is they are a big, massive,

42:37

adaptive boundary that has been

42:39

constructed between the termites themselves and the

42:41

environment, which

42:43

they are, of course, totally unsuited to be living

42:46

on their own. And the more

42:48

I studied them, the more I

42:50

came away impressed with just

42:52

how extensive this reach was, you know. So

42:55

it extends not only to managing the

42:59

atmospheric composition within the nest, but

43:01

it also co-ops

43:04

the physical environment, the entire

43:06

hydrology of the environment over

43:08

a fairly extensive range

43:10

to be able to

43:12

enable termites to live in a

43:15

dry environment, but because they

43:17

reconstruct their environment to manage water flow through

43:19

it, they can survive in those

43:22

kinds of environments. Cortescott's thesis

43:24

is that organisms and perhaps other levels

43:26

of biological organization have to be intentional.

43:29

Here he talks about how mole crickets actively

43:32

modify their burrows to get just

43:34

the sound they want so they can attract mates. In

43:36

some cases, like in humans, consciousness

43:38

can come into play and be an aspect

43:40

of intentionality. But intentional

43:42

behavior need not be conscious. It

43:45

just needs to be directed at something in the environment based

43:47

on some pre-existing model in

43:49

the brain or elsewhere in the body about expected

43:52

outcomes and current needs. Again,

43:54

states of a Markov blanket.

43:56

So what is it that we do

43:59

when we...

44:00

have an intention. You know, well,

44:02

we, there's a conscious part

44:04

of it, definitely. Kind of want to stay away from

44:07

that a little bit, but, you

44:09

know, this intentionality

44:12

can be framed

44:13

in a way that links the cognitive

44:15

interpretation of the environment with the connection

44:18

to the engines,

44:20

if you will, that can modify the environment.

44:23

And so, when you look

44:25

at the burrows

44:28

of, the tuned burrows of mall

44:30

crickets, for example, you know, these

44:35

creatures build a burrow, it ends

44:37

up in the shape of an exponential horn, this

44:39

helps project the

44:41

sound of the call much further than

44:43

it would otherwise. And if you look at what's

44:46

happening during the construction of that burrow,

44:48

the cricket burrows a little

44:50

bit, emits a chirp,

44:53

listens to it, and if it's not quite right,

44:55

it continues to modify its burrow

44:58

until it gets the chirp that it wants.

45:00

Again, I'm putting up scare quotes here, that

45:02

it wants. Yeah, right. And

45:05

that's kind of an intentionality, isn't it?

45:08

And so, if we want to try to develop

45:10

a concept of what intentionality is

45:13

that can be kept

45:15

independent from the kind

45:17

of mysticism that

45:20

tends to trip this up, then to

45:22

me, the simplest

45:23

definition is coupling

45:27

modification of the environment with the

45:29

cognitive interpretation of the environment.

45:40

Before wrapping our story, we think it's important

45:42

to point out the practical reasons for understanding

45:44

agency. One more time, here's Mike Levin on

45:47

just two such reasons. First,

45:49

our health. So imagine in the next 10 years,

45:51

we solved two things. We're

45:53

going to solve genome editing. So somebody

45:55

will have come up with a nice clean way

45:58

of making genetic edits.

45:59

where you want it and nowhere else. For perfect editing.

46:02

Yeah, yeah, yeah. Forget all that. Let's say you

46:04

get totally perfect editing. And let's say, and by

46:06

the way, let's also say that stem cell biology

46:08

gets solved so that from a stem cell, you

46:10

can get any other single cell type that you want. Okay.

46:13

So now you have all this. And

46:15

so now that means that you're going to solve some

46:17

really nice low hanging fruit. So

46:20

single gene diseases of which there are

46:22

some and single cell

46:24

diseases, you know, Parkinson's maybe things

46:27

like that.

46:28

But then, then you're going to reach

46:31

the much, much deeper question. Okay. Somebody's

46:33

missing their hand. Let's say there was an accident or birth defect

46:36

or whatever or an eye.

46:38

And now what? Because the point

46:41

isn't to be able to edit the genome cleanly.

46:43

The point is what in the world would you edit? So

46:45

if you don't believe, you know, if we

46:47

don't have a good account of

46:49

morphogenetic agency

46:52

and competency, if we don't have

46:53

a reverse engineering of this kind of software

46:56

of life that enables it to have modularity

46:58

and so on,

46:59

you're talking about micromanaging at the

47:01

molecular level,

47:03

all of the steps that go on to making a complex

47:05

organ. That is not going to happen. Nevermind

47:07

our lifetime. You know, I don't know how many years it's going to be before

47:09

that's even feasible. If it even is at all feasible,

47:12

that kind of micromanagement. Second, here's Mike

47:14

on rapid advances in our technology.

47:16

We are going to see cyborgs

47:18

and hybrids and, you know,

47:21

every kind of combination of biology,

47:23

technology, artificial intelligence and

47:25

software and hardware merges

47:28

of, of living tissue with, with

47:30

engineered, you know, all kinds of things. What

47:32

that means is the older categories,

47:35

things like

47:36

what is a robot? What is a machine?

47:39

How do we recognize agency?

47:41

What is, what is something that was evolved

47:43

versus design? Does it matter for

47:46

these things? We have to start wrestling

47:49

with this now because in the olden

47:51

It was very easy to tell.

47:54

And even then, of course, we made all kinds of

47:56

mistakes with various kinds of humans and

47:58

animals. We made all sorts of terrible mistakes.

47:59

But generally speaking, you

48:02

could do this. You would sort of knock on something.

48:04

And if you hear a metallic sound, you would say, oh

48:06

yeah, you can do whatever you want with this. And if

48:08

it was squishy and sort of warm

48:10

and furry, you would say, if

48:13

you do certain things with this, you're going to jail,

48:15

right? You have to be nice to this one. It's a

48:17

horse or a dog, whatever. And

48:19

that was easy because you could rely on two

48:21

things. You could rely on the thing

48:23

it's made, what it's made of.

48:25

And you could rely on an origin story. You

48:27

could say, well, this thing evolved.

48:29

And this thing was created in the lab. And that makes

48:32

all the difference. Those categories are gone.

48:34

I think even now these categories are no good.

48:36

And they're absolutely going to be no good going forward.

49:05

Let's suppose that Darwin had remained in the

49:07

clergy, that Fisher failed out of math,

49:09

and Mendel flew kites instead of growing peas. Suppose

49:11

that Lamarck, Shannon, Cannon, Waddington,

49:14

and McClintock got more positive attention

49:16

or just weren't excluded by others in their

49:18

fields. Suppose Lysenko was honest or

49:20

was just ignored. Suppose that Williams, Franklin,

49:23

Watson, and Crick had access to the computing

49:25

power of the average smartphone. Or

49:27

suppose that Dennis Noble, Yuri alone, or other

49:29

systems biologists had been working in the early

49:31

1900s. Suppose that biomedicine

49:34

hadn't become a barotting capitalistic behemoth.

49:36

If these counterfactuals were true,

49:38

biology today would probably be a discipline much

49:41

less focused on genes and much more

49:43

focused on life, is what Stuart Newman

49:45

calls active matter. Let's make

49:47

it so.

50:09

And

50:09

before we go, as promised, here's a quick

50:11

exchange between Cam, Marty, and me on agency.

50:14

Clearly Cam views agency differently than

50:15

we do.

50:19

Okay, we've got the three of us here in a room. It's

50:22

me, Marty, and Cam, and we're going to talk

50:24

for a little while about agency. We have a divergent

50:26

set of points of view about the importance

50:29

and utility of agency and sort

50:31

of bigger or broader issues in biology,

50:33

like how much does

50:35

the modern synthesis need updating. So,

50:38

Cam, why don't you go first and

50:40

have at it. So I want to start off

50:42

by saying that I don't deny that

50:45

organisms are complex systems

50:47

and that agency

50:49

exists. I absolutely

50:52

agree that living systems have evolved to

50:54

be robust, to

50:56

exhibit homeostasis, to

50:58

be flexible, to be plastic,

51:01

to be self-regulating, that

51:03

they interact with their environments in

51:05

complex ways. And

51:08

you can see this at different sort of

51:10

levels of biological organization.

51:13

But what I'm struggling with is, I guess

51:15

like my first question for you is, can

51:17

you have agency

51:20

without natural selection?

51:22

Yeah, sure. Can

51:24

you give me an example of that? Okay,

51:27

so this is where it gets interesting and yet

51:29

maybe complicated and off page. Immediately.

51:32

Yeah, we just first question and it's already off

51:34

the rails.

51:36

By natural selection, I'm assuming that you mean

51:38

the one that Darwin pointed to,

51:40

the one that Darwin popularized,

51:42

or you mean something broader.

51:44

Is there another version of natural selection

51:47

aside from the one that Darwin coined? So

51:50

natural selection, what I'm saying is

51:52

that evolution can happen independent of life.

51:55

None of us would argue that, I think.

51:57

You don't have to have living systems to evolve.

51:59

to have complex systems that

52:02

change through time, one of which is life.

52:04

So Darwin didn't bother to talk about any other kind

52:06

of complex system. He was only interested in the one with

52:08

fins and feathers and such.

52:10

But Marty, I think I may disagree with

52:12

you on this point. So do you think you can

52:15

have agency without selection

52:17

no matter how you define a selection? Like to me, it doesn't actually

52:19

really matter how we define selection. Let's just

52:22

take it as

52:23

Darwin's kind of natural selection and

52:25

then try to use that to answer Cam's question.

52:28

Can you get agency without that

52:30

form of selection operating? I

52:33

don't know that I can answer the question that way because I have it

52:35

in my head in such a different way. Let me just

52:37

try really briefly to articulate what it is

52:39

that I mean and why I'm pushing back on the natural selection

52:42

that's not Darwinian.

52:43

You're inevitably, any time you get at the existence

52:45

of a complex system, where you get a system that

52:48

comes to sustain itself through time,

52:51

the process by which that happens,

52:53

as long as sustaining happens for long enough, if you ever

52:55

get to a point of replication, you're gonna have the instantiation

52:58

of information such that the future generation to

53:00

that system are different than the last ones. That's

53:02

evolution by natural selection of any system living

53:05

and non-living. That's what I'm talking about.

53:07

So Darwin happened to pick

53:09

on

53:10

living systems,

53:12

but that process

53:14

should apply to any kind of system

53:16

that persists in time.

53:18

So you would say, for example, that like, we

53:21

talked to Tim Linton about this, and he

53:23

had this idea of say, grasslands

53:25

being a complex system, that

53:27

have a lot of different components that are interacting in

53:29

the grasslands, because of the

53:32

way grasses affect fire dynamics and grazing

53:34

dynamics, and the interaction between

53:36

grasses and trees, those systems are

53:38

complex, and they're self-sustaining

53:41

over long periods of time. So

53:43

would you say that that ecosystem

53:45

has agency?

53:47

I, yeah, I have a hard time with that. I

53:49

think just because of my bias about organisms,

53:51

based on everything that I've said, and most of

53:54

the people that we talked to, including Tim, I think they

53:56

would say yes, that has agency.

53:58

Because agency in the most...

53:59

generic sense is the Carl

54:02

Friston one

54:04

of an updatable set of Markov blanket

54:06

states, right? And so as long

54:08

as you have one that reifies

54:10

itself because of the way that it's updated,

54:12

it states that should mean that it has agency.

54:15

I suppose I could accept that, although it

54:17

seems a lot more likely to me that it's,

54:19

you know, individual organisms and

54:22

parts within those organisms that are going to actually

54:24

develop sophisticated forms

54:26

of agency

54:27

because that agency itself is

54:29

going to be shaped by natural selection.

54:32

Well, yeah, now natural selection, you mean a Darwinian

54:34

one?

54:35

I do. Yeah. So, I mean,

54:37

I agree with you, but I'm uncomfortable

54:40

and fully agreeing with you only because I can't get

54:42

my head around why that should fundamentally be different,

54:45

organismally and suborganismally than

54:48

at the level of what Tim was talking about.

54:50

Well, I'm still

54:52

a little bit confused by

54:54

non-Darwinian selection.

54:56

I mean, I think in the general

54:58

sense, when we talk about evolution

55:01

by natural selection, we're referring

55:04

to a

55:05

set of conditions that when those

55:07

conditions are met, then there

55:09

is some predictable outcome. And those

55:11

conditions are simply just that there's variation.

55:14

Some of that variation is heritable. And

55:18

if some of that variation is associated

55:21

with differential survival or

55:24

reproduction,

55:25

then those individuals

55:28

that, you know, have higher fitness become

55:30

more represented in a population.

55:33

And so in

55:34

the most general sense, that

55:37

can apply, you know, if we think about

55:39

it as a levels of selection problem,

55:42

something like a transposable element

55:44

is a selfish

55:47

bit of DNA. It doesn't have

55:49

like a heritable component in

55:51

the same way that a multicellular

55:53

organism would have. But

55:56

if there is an

55:58

element that tries to propagate

56:00

and make more copies of itself, then

56:03

in within the environment

56:05

that it lives in inside that genome, it

56:07

has higher fitness, and

56:09

it will propagate and increase.

56:12

That's

56:12

what happens. That's true.

56:15

But I think what we're talking about is a different kind

56:18

of thing. It's a much more inclusive thing

56:20

than the sort of temporal

56:23

changes in lineages like transposable

56:25

elements. Well, but I guess

56:28

that still confuses me because

56:31

you could think of a computer program as also

56:33

something that could evolve. It has some sort

56:36

of information. It's not living,

56:38

but it still conforms to

56:40

the same principles of

56:44

Darwinian natural selection.

56:45

Even though it's not a living

56:48

organism, it's still those

56:51

programs that have higher fitness

56:54

increase, those that have lower fitness

56:57

decrease. It's

56:59

the same general concept. That's why I guess

57:01

I'm struggling with

57:02

why that's different from

57:04

any other kind

57:07

of system that might evolve.

57:09

There's two big things in

57:12

the example that

57:14

you're using that stick out for me. One of them is

57:16

that when we say

57:18

that evolution by natural selection from

57:20

Darwin is those big three, heritability,

57:23

variation, and differential survival

57:25

and reproduction. That's all true,

57:27

but

57:28

it doesn't drive home what

57:31

Scott Turner really emphasized, where

57:33

Darwin was really coming from and a lot of people around

57:35

his time. It doesn't capture the

57:37

struggle for existence component

57:39

nearly well enough, meaning that it over

57:42

simplifies

57:43

how hard it is to be alive. Before

57:45

you even get to the heritability and variation, it's

57:48

just plain old, what is it to be

57:50

alive that will allow any of those other three

57:52

things to make a difference? The second

57:54

piece that I think is really important in this same

57:56

space

57:57

is why people traditionally pick on...

58:00

traditional thinking of evolutionary biology, we don't

58:02

have ideas about the origins of variation.

58:05

We just say that there's variation,

58:06

but coming from an agential homeostatic

58:09

complex systems mindset, immediately

58:12

for free, when you start talking about minimizing

58:14

surprise and entropy reproduction or entropy

58:16

exporting,

58:17

you end up with an explanation for not just

58:19

how much variation, but the sources and

58:22

kinds and variety of information that so

58:24

many people wanna know. So it's just

58:26

a richer landscape.

58:28

I mean, it just does work that isn't

58:30

offered by modern theory,

58:32

but I'd be interested to hear what Art says about that.

58:34

I think I agree with what you just

58:36

said. I was hung

58:39

up on thinking about the

58:41

computer software program that Cam was just

58:44

mentioning. And I was thinking maybe of a similar

58:46

sort of analogy last night when I was anticipating

58:49

this conversation. And I was

58:51

trying to think about like, what

58:53

would be a fair way to characterize

58:56

the difference between maybe

58:57

a modern synthesis view

59:00

and something that took a broader stance.

59:03

And this comes from, I think the comments

59:05

that you've made, Cam, about this

59:08

conversation in our writings to each other and

59:10

what I've heard from other people.

59:12

And that is, I think you could say, well, this

59:15

agency stuff is all really interesting and

59:18

yeah, physiology interesting, but that's

59:20

not what the modern synthesis was designed

59:23

to do. It was designed to provide a very

59:25

simplified quantitative statistical way of

59:27

understanding how variation and

59:30

filters on that variation is translated into

59:32

micro and macro evolutionary change over time.

59:36

And it does that quite well.

59:38

But okay, so here's this analogy that I had in mind

59:40

last night thinking about this. So imagine

59:43

we're trying to explain the evolution of flying

59:45

machines from say the Wright brothers plane

59:47

up to

59:48

modern fancy jets. What

59:51

would the modern synthesis say about that

59:54

sort of evolution and diversification

59:56

of flying machines? It would be something about

59:58

the plans,

1:00:01

the blueprints, the electronics

1:00:04

that have diversified and become

1:00:06

more complex over time.

1:00:09

And we could even draw phylogenies

1:00:11

of airplanes that were based on what we know

1:00:14

about how that information was transmitted

1:00:16

among individuals and among companies.

1:00:19

But that knowing that somehow still

1:00:22

doesn't explain that much

1:00:24

about how airplanes operate,

1:00:26

about

1:00:27

where the Bernoulli effect comes from,

1:00:29

about why the wings and the tails

1:00:31

are where they are and how you steer the plane in

1:00:34

the air. Like all this interesting stuff

1:00:36

about

1:00:37

what makes a plane a plane isn't captured

1:00:39

by that sort of theory of transmission

1:00:42

of plans. And so I mean

1:00:44

maybe this is not a good example because like

1:00:48

what I'm invoking here is not like different parts of an

1:00:50

airplane that have agency so

1:00:52

much as saying you

1:00:54

know there's this sort of narrow path that describes

1:00:56

the evolution, but we want to know a lot

1:00:58

more than that

1:00:59

because that's where a lot of the interest lies.

1:01:02

So I agree with you and I think

1:01:05

this is a little bit reminiscent of

1:01:07

some

1:01:08

debates that happened sort of

1:01:10

at the interface of ecology and evolution

1:01:13

between sort of the evolutionary

1:01:15

explanations for patterns

1:01:18

versus sort of more proximate

1:01:21

mechanistic explanations. And

1:01:23

you know during the 50s and 60s

1:01:26

I think there was a lot of debate about just

1:01:28

among ecologists to try to explain

1:01:31

the phenomenon that they saw. It

1:01:33

was very strictly a mechanistic

1:01:36

interpretation of what regulated

1:01:39

populations you know through for

1:01:41

example rainfall and they

1:01:43

viewed those explanations as very distinct

1:01:46

from any kind of evolutionary explanation

1:01:48

that like ecology and especially

1:01:51

sort of like functional ecology, ott ecology

1:01:54

was its own separate discipline and provided

1:01:56

its own set of explanations

1:01:59

for how things work.

1:01:59

And you can think about it in the same sense

1:02:02

of like a more mechanistic explanation

1:02:04

for behavior. You know, an organism

1:02:07

exhibits a certain behavior because there's

1:02:09

a stimulus from the environment

1:02:11

that causes a hormonal response,

1:02:14

which triggers some sort of neural response,

1:02:17

which then, you know, eventually through

1:02:19

various complex pathways results

1:02:22

in a certain kind of behavior.

1:02:24

Those are not evolutionary explanations.

1:02:27

The evolutionary explanation for that would

1:02:29

be that, you know, the animal does that behavior

1:02:31

because it gives it a fitness advantage.

1:02:34

It doesn't say anything about the nuts and bolts

1:02:36

about that. And, and I think there was a lot

1:02:38

of debate about this through the fifties

1:02:40

and sixties and Ernst Meijer wrote

1:02:43

about this kind of cause and effect problem

1:02:45

in biology and, and in the end,

1:02:48

you know, the consensus was that, look,

1:02:50

we're studying the same things and these

1:02:52

approaches are not antagonistic

1:02:54

to one another,

1:02:55

they're complementary to one another. And

1:02:57

we shouldn't think of them as being opposed. We

1:03:00

should think of them as, you know, different

1:03:02

ways of understanding.

1:03:04

And so when reading Ernst

1:03:06

Meijer's descriptions of the

1:03:08

big meeting that sort of led to

1:03:11

the coining of the term modern

1:03:13

synthesis, he, he makes some comment

1:03:15

about how they invited developmental biologists

1:03:19

and physiologists to attend the meeting and nobody

1:03:21

was interested.

1:03:22

It was because they really didn't see

1:03:24

how they thought they were unrelated.

1:03:26

Yeah. They, it just, it didn't cross their minds

1:03:29

that, you know, what you study at

1:03:31

an organismal level, the mechanisms of

1:03:33

physiology and development and behavior would

1:03:36

have anything to do with, with these sort

1:03:39

of population level, evolutionary responses.

1:03:42

And so I, I wonder if

1:03:44

some of the

1:03:46

debate and disagreement over

1:03:48

the importance of agency and, and

1:03:50

some of these other mechanisms, like understanding

1:03:52

the nuts and bolts of like an airplane,

1:03:55

the physics and the mechanics and the electronics

1:03:58

are like, those are.

1:03:59

those are also very important and those are very

1:04:02

complimentary, but those aren't

1:04:04

going to be explained by a general

1:04:07

theory of population genetics

1:04:10

that is thinking about the processes

1:04:13

of genetic drift and mutation. And

1:04:15

those all become, I think what Marty

1:04:18

calls the vicious abstraction. You

1:04:20

can't see those mechanisms,

1:04:23

but you have to sacrifice. And

1:04:25

I think that one thing that

1:04:27

I know both you and Marty have talked

1:04:30

a lot about as kind of a different perspective

1:04:32

is the systems level perspective.

1:04:35

And I think systems biology

1:04:38

and systems level thinking, that

1:04:40

does offer a different perspective

1:04:43

on the sort of traditional

1:04:45

single locus, two allele

1:04:47

population genetic kind of simplification

1:04:50

model because it does bring in a lot

1:04:52

of complexity that isn't

1:04:54

necessarily in those population genetic

1:04:57

models, but people are studying

1:04:59

that. There are evolutionary biologists

1:05:01

that think about, for

1:05:03

example, the concept

1:05:06

of robustness. Robustness is not

1:05:08

a concept that is easily pulled

1:05:11

out of quantitative genetics or population

1:05:13

genetics, but it's still a type

1:05:16

of problem and phenomenon that's

1:05:18

studied within evolutionary biology.

1:05:20

I just first wanted to make a

1:05:23

comment about this idea of vicious

1:05:25

abstraction, which we talked a lot about in the

1:05:28

script that we just read and that you just brought up. And

1:05:30

to be clear, there's nothing wrong with

1:05:33

simplification in models. And in fact,

1:05:35

that gives them

1:05:36

enormous amounts of power. I mean,

1:05:39

that's essentially the philosophy of doing a model

1:05:41

is what's the minimal amount of stuff you can

1:05:43

write down to capture something that's important

1:05:45

and of essence in a system. And

1:05:49

another non-evolutionary example of that would

1:05:51

be thinking about, well, what are good

1:05:53

mathematical models of how populations fluctuate

1:05:55

over time? And there's the exponential

1:05:58

growth model, which has just... a very

1:06:00

simple equation that underlies it. There's the

1:06:02

logistic growth equation, which adds

1:06:04

in just another couple of terms

1:06:07

that allow you to incorporate density

1:06:09

dependence into an exponential

1:06:12

growth model. And those have been vastly powerful

1:06:15

precisely because they're really vicious abstractions,

1:06:18

right? And so I think the

1:06:20

way to relate that back to what you're just talking about,

1:06:22

about evolutionary theory, is that the

1:06:24

modern synthesis is based on these abstractions

1:06:26

that are super powerful because they boil it

1:06:28

down to some kind of essence that really

1:06:29

matters. And I

1:06:32

think one path would be to say,

1:06:35

this stuff about agency and physiology

1:06:37

and homeostasis is just something else.

1:06:40

And it's not what the modern synthesis

1:06:42

was designed to explain. And

1:06:44

the problem with that, I think, is that what I

1:06:47

can see glimpses of is that

1:06:49

this complex system stuff and agency

1:06:51

and homeostasis feel to me like they feed

1:06:54

back on that evolutionary process

1:06:56

in a really important way. In other words,

1:06:58

by doing the vicious abstraction, you've

1:07:00

actually lost sight of something that's

1:07:03

not tangential, but something that's really central

1:07:05

to the way lineages evolve

1:07:07

and diversify. And so it feels

1:07:10

like we need to bring that back into the main thread.

1:07:14

Yeah, that's where I would end. Because first,

1:07:16

to be fair and clear, that's not my word. I'm

1:07:18

not that creative. That was William James' particular

1:07:21

word, but it wasn't him that claimed that necessarily.

1:07:24

Vicious abstraction. Yeah, he was making a general

1:07:26

point about, when you're trying to model

1:07:28

something by its nature, you want to simplify it

1:07:30

with defeat. But I think it's at

1:07:33

core. It's been really good,

1:07:35

really generous of you to include all of this text

1:07:37

and the script and the other emails and things, Cam, because I think

1:07:39

at the bottom, we're

1:07:41

interested in different

1:07:43

things. It's not surprising to me for

1:07:45

you to say that you believe that agency

1:07:48

is real and all those sorts of things. And I think most

1:07:50

evolutionary biologists, not everyone, but

1:07:53

most of them would go along with

1:07:55

agency if agency is defined

1:07:57

in a really specific way as to be. the

1:08:00

kinds of things that help organisms not fall apart

1:08:02

or help systems not fall apart.

1:08:04

But I think that

1:08:06

the piece that's a little bit weird, what Art

1:08:08

is alluding to, it's not so much that the modern

1:08:10

synthesis was wrong. The question

1:08:12

is whether the original vicious abstraction

1:08:15

is the best, most

1:08:17

effective abstraction. And

1:08:19

I just think that it can't be exactly what

1:08:21

replaces it. Now, that's a totally fair

1:08:23

question. And I think, you know, when the people

1:08:25

are kicking the modern synthesis and people saying, what else

1:08:27

do you want? Just listing terms, niche

1:08:30

construction or epigenetics, that's not enough, because

1:08:32

that's just the laundry list. It's something else that isn't

1:08:34

necessarily incompatible. So

1:08:37

it is the responsibility of those that want

1:08:39

something different to articulate what's different. And

1:08:41

I think that's you know, that's why we did this thing with agency.

1:08:44

It seems to be because most

1:08:46

of this stuff are like they're all systems. It's

1:08:48

really interesting to me to hear you

1:08:51

say that you are cool with robustness

1:08:54

and systems thinking, but since all of this other

1:08:56

stuff gets a little bit wonky, you're drawing

1:08:58

the line in a different place than I can under I can

1:09:00

understand, because I don't think it's really all that different. But

1:09:03

but at the end of the day, wouldn't you say that we want we

1:09:06

all want a vicious abstraction, we just

1:09:08

want a different one?

1:09:09

Or is there value in a different one? Well, I

1:09:11

think that any kind

1:09:14

of model that

1:09:16

improves our understanding

1:09:19

of how things work is something that, you

1:09:21

know, who wouldn't want that everybody wants

1:09:23

that. And I think it's

1:09:26

I think that, you know, my problem

1:09:28

with a lot of the

1:09:30

people who, you know, beat up

1:09:32

the standard evolutionary theory and say

1:09:34

like, well, it doesn't include epigenetics,

1:09:37

or it doesn't include niche construction,

1:09:40

it doesn't include plasticity, even

1:09:42

though I study plasticity, and I find

1:09:44

plasticity perfectly compatible within

1:09:47

the context of, you

1:09:49

know, the standard evolutionary theory, what

1:09:52

I haven't seen articulated is

1:09:55

what changes. So at the

1:09:57

core of like what we would

1:09:59

call standard evolutionary theory are

1:10:01

certain processes. You

1:10:04

know, those include mutation,

1:10:06

genetic drift, gene flow,

1:10:09

recombination selection. And,

1:10:13

and then you can, you can study and you

1:10:15

can model and you can empirically measure

1:10:18

interplay between gene flow and selection.

1:10:21

Well, let me, let me, let me add

1:10:23

something else in here, because I guess, I think I didn't finish

1:10:25

the thought before it's, it's not

1:10:28

just that we want a model that includes

1:10:30

a vicious abstraction, because useful models have

1:10:32

to have vicious abstractions.

1:10:34

What is it that we're trying to model? What

1:10:36

I'm talking about is not the same

1:10:39

thing because modeling evolutionary change,

1:10:41

especially when the history of those kinds of models have

1:10:43

been about genetic

1:10:46

change,

1:10:47

right? This kind of mentality,

1:10:49

whatever the suspicious abstraction is going to be, will

1:10:51

be something about the viability,

1:10:54

the sort of sustainability of a system.

1:10:56

Now that can include

1:10:58

reproduction, right? And evolution

1:11:00

subsequently, but it doesn't have

1:11:02

to. So we're talking, we're coming up with a model

1:11:05

that explains the existence,

1:11:07

like the origins of life, the persistence

1:11:10

of life and variations into

1:11:12

the future on life based on

1:11:14

these markup updates and that kind of thing. But

1:11:16

it's not quite the same thing. So, you know, gene

1:11:18

flow and drift and those kinds

1:11:21

of things aren't necessarily

1:11:23

even included. And why I'm picking

1:11:25

on this, and I want to hear what you think about it, knowing

1:11:27

that this is the like, to what end are we modeling?

1:11:30

I

1:11:30

just really strongly feel that

1:11:32

even though the modern synthesis was not

1:11:35

intended to

1:11:36

have the impact on biology

1:11:38

that it has had,

1:11:40

it's time to confront the fact that it has,

1:11:43

that biomedicine, for example,

1:11:45

spends an enormous amount of money, so

1:11:47

much more money on things genetic

1:11:50

than things other,

1:11:51

right? And so practically, it's

1:11:53

time to stop living

1:11:55

in the modern synthesis world, or

1:11:58

tacitly representing it

1:12:00

as our backbone because

1:12:02

people's lives are at stake.

1:12:15

So I think part of my disagreement

1:12:17

then is holding up this

1:12:20

thing that people keep referring

1:12:22

to the modern synthesis, which

1:12:25

is, you know, it's like by it's

1:12:27

like evolutionary biology had the synthesis 50

1:12:30

years ago and nothing's happened ever since, which

1:12:33

is how it's often depicted. That's

1:12:36

unfair. That's you're right. So I think

1:12:38

a more a more productive maybe

1:12:41

way is like within kind

1:12:43

of what we would think of a standard evolutionary

1:12:46

theory, what are the explanations

1:12:48

for what I think your

1:12:50

first step that you're very interested

1:12:53

in is

1:12:53

the origin of life. Like

1:12:56

I know, for example, you've interviewed people

1:12:58

like Sarah Walker and Nick

1:13:00

Lane and you know, does RNA

1:13:02

come first or does metabolism come first?

1:13:06

The population genetic based

1:13:08

theory doesn't say anything about that. There

1:13:11

may be some other theory out there

1:13:14

that maybe invokes a role for natural

1:13:16

selection in favoring

1:13:19

one variant over another

1:13:21

variant of you know, which which

1:13:24

one may be more effective, but but

1:13:26

I think that's that's more kind

1:13:28

of a discussion for

1:13:31

biochemists and

1:13:34

chemists and and biologists.

1:13:36

This well, it you know, it's

1:13:38

certainly more at the interface. And

1:13:41

and so so then if we if we want to

1:13:43

then move beyond that to say that, you know,

1:13:45

successful systems are those

1:13:48

that

1:13:48

do a good job of replicating themselves

1:13:51

that can keep themselves

1:13:53

alive and persist entropy

1:13:56

and maybe modify their

1:13:58

environments in way that

1:14:00

ways that make them

1:14:01

survive better and more

1:14:04

suitable for themselves.

1:14:05

I think that's all fine. I think articulating

1:14:09

a model or

1:14:12

a theory for that

1:14:14

kind of process of life that's very,

1:14:17

very general.

1:14:18

That is totally fine. And I

1:14:20

think standard evolutionary theory

1:14:23

maybe can be incorporated into that to

1:14:25

talk about, you know,

1:14:27

how things go in one direction

1:14:29

versus the other or...

1:14:31

But I don't, I don't see, I see

1:14:33

that as a,

1:14:35

again, more of a complementary

1:14:37

type of theory. It's

1:14:39

maybe most evolutionary

1:14:41

biologists are focused on micro

1:14:44

evolutionary change, but, you know,

1:14:46

I can say going to like the evolution meetings,

1:14:49

you don't see people giving talks, for example,

1:14:51

or presenting posters on...

1:14:54

At least it's not very common on like, did

1:14:56

RNA come first or did metabolism

1:14:59

come first? Like, certainly that's

1:15:01

a big evolutionary question, especially,

1:15:04

you know, in the history of life.

1:15:05

But that at bottom is the thing.

1:15:08

There's a bunch of people, I think, Art

1:15:11

and I are really physiologically inclined.

1:15:13

I think we get jazzed about this

1:15:15

because the number one thing for us

1:15:17

when you talk about any living system

1:15:19

is that it's not dead. That there has to

1:15:21

be some on these days. I mean, it's... I'm

1:15:24

not dead yet. But that's the core,

1:15:26

not dead yet. But that's the core. And evolutionary

1:15:28

biologists,

1:15:30

to me, bizarrely, and I am one because my PhD

1:15:32

says it, but they don't care necessarily

1:15:36

that life resides in those equations.

1:15:39

And that's just perplexing.

1:15:41

And I don't think, and I don't think even viable in 2023.

1:15:44

But Art, what do you think? You've been quiet. Well,

1:15:46

I want to go a different direction. So what

1:15:48

I hear you guys arguing about is,

1:15:51

you know, whether we should localize this argument

1:15:54

onto a sort of more standard

1:15:56

view of how we view evolution right now among

1:15:59

living organisms. using the mechanisms that we understand,

1:16:02

versus taking this approach that Marty

1:16:04

is advocating that sort of goes beyond

1:16:08

evolution of individual populations

1:16:11

and, you know, genes and differential survival.

1:16:13

So to sort of thinking about complex

1:16:15

systems and their origin more broadly.

1:16:17

I want

1:16:18

to circle back to sort of how I feel

1:16:20

like some of these ideas affect my

1:16:23

thinking about more standard use of evolution.

1:16:25

And this comes out of just a couple days

1:16:27

ago. So Alicia Shaw

1:16:29

invited me to Kellogg Biological Station.

1:16:31

I gave a talk there a couple of days ago and did

1:16:33

a lot of thinking about how all of these

1:16:35

ideas interface with my own work on thermal

1:16:38

ecology and thermal physiology of insects.

1:16:41

And

1:16:42

I was struck again by, I think it's this

1:16:44

idea that maybe was articulated most strongly by Scott

1:16:47

Turner, that what's interesting about agential

1:16:50

thinking is that if you

1:16:52

think about how populations evolved,

1:16:54

you know, what is there? There's some kind of variation,

1:16:57

there's some kind of filter, and

1:16:59

then some subset of that

1:17:01

original variation makes it through that filter

1:17:04

and becomes the subsequent population

1:17:07

that's reproducing. And that's a very kind of micro-revolutionary

1:17:10

view. It's a standard thing that we all say in our

1:17:12

basic biology classes. And what agency

1:17:15

adds to that is that it means

1:17:17

that the organisms that have the variation

1:17:20

are also by their actions,

1:17:23

you know, creating and modifying the

1:17:25

very filter that is selecting

1:17:28

on them. And that to me feels like,

1:17:30

you

1:17:31

know, you could say that's a small thing, and we could

1:17:33

explain that as, you know, the fact we've known

1:17:35

forever that,

1:17:36

yeah, organisms interact with their environment, organisms

1:17:38

have behavior. But

1:17:40

I think that underplays this sort of

1:17:42

fundamental importance of organisms both,

1:17:44

you know, at some level creating the variation

1:17:47

and modifying and creating the very

1:17:50

filter that's doing that selection.

1:17:52

And that feels to me like a sort

1:17:54

of profound shift in thinking about where

1:17:57

variation

1:17:59

and

1:17:59

you know, potential causes of change

1:18:02

come from, discuss.

1:18:03

Yeah, so

1:18:06

I think that is a

1:18:09

more sort of rich and

1:18:12

productive way of thinking about things. But

1:18:14

again, that kind of thinking is also

1:18:17

captured, for example, within the,

1:18:20

what people call eco-evolutionary dynamics.

1:18:23

So, like the guppies that

1:18:25

I study, work by Dave Resnick,

1:18:27

John Endler and others, but Ron

1:18:30

Basser and Joe Travis recently

1:18:33

have shown, for example, that

1:18:35

when you move guppies from a river

1:18:39

that is full of predators, that

1:18:41

keeps the population density of the guppies

1:18:43

down, you get a certain life history

1:18:46

that evolves, sort of a live fast,

1:18:48

die young kind of strategy. And

1:18:51

whether guppies naturally colonize

1:18:53

or experimentally are put into

1:18:55

these streams where these predators are absent,

1:18:58

they modify their environment,

1:19:00

but not always in a good way.

1:19:02

In the absence of predators,

1:19:04

the populations grow, they

1:19:07

become, you

1:19:08

know, very high density

1:19:10

kind of populations. And you would

1:19:12

think that maybe in the absence of predators, this

1:19:15

is sort of a nice,

1:19:17

happy paradise. But in fact, it's

1:19:19

kind of a nasty place to live. You're

1:19:21

in these very small streams under a closed

1:19:24

canopy with very little productivity, and

1:19:26

the population densities are high. So

1:19:29

it's a very competitive, very nasty

1:19:31

kind of environment. And the sort

1:19:33

of previous thinking was that the evolution

1:19:36

of the sort of low, the

1:19:38

more slow life history

1:19:40

was simply just due to the absence

1:19:43

of mortality from predators. But

1:19:45

now it's appreciated that it's

1:19:47

much more driven by the density

1:19:49

dependent

1:19:51

kind of competition for food

1:19:53

and the sort of

1:19:55

nastiness that occurs in these kinds

1:19:57

of streams. So

1:19:59

And of course, that then changes how

1:20:02

the Guppies evolve, right? They

1:20:04

go from one situation to

1:20:06

the other. And I think that's where

1:20:09

the concept of agency, I

1:20:11

worry about that then becoming a bit circular.

1:20:13

Because if Guppies weren't

1:20:17

able to adapt and evolve to these new

1:20:19

set of conditions, they would go extinct. And

1:20:21

the fact that they don't go extinct means that

1:20:24

there's been some sort of compensatory evolutionary

1:20:27

change in the life history and

1:20:29

the physiology and the behavior that allows

1:20:31

them to persist.

1:20:32

And so the system, in this case,

1:20:35

the population level, persists in

1:20:37

this new environment. And it's been shown

1:20:39

that this can evolve very quickly.

1:20:42

And it's been replicated many, many times.

1:20:45

So is that kind of interaction

1:20:47

between the Guppy and its environment

1:20:50

agential? I would refer to that

1:20:52

as either density dependent evolution

1:20:54

or ecoevolutionary dynamics. We

1:20:57

could also call it agential behavior.

1:21:00

But I think all three are capturing

1:21:02

the same kind of interaction

1:21:05

between the

1:21:06

organism, the environment, changing

1:21:08

the selection pressure,

1:21:10

acting on that variation, and

1:21:12

this more kind of feedback loop,

1:21:15

rather maybe a more linear sort

1:21:17

of response. Sure, sure. I

1:21:19

mean, I think honestly, that's a super beautiful example.

1:21:22

And to me, that's a great example

1:21:25

of organisms creating the conditions

1:21:28

themselves that are then selecting back

1:21:30

on them. And density dependence is clearly an

1:21:32

example of that that happens not only in Guppies, but in

1:21:34

many other groups. But

1:21:37

to me, what's important about this idea of

1:21:39

agency

1:21:40

is that it allows me to see that

1:21:42

that kind of density dependent selection that you

1:21:44

just described is a subset

1:21:47

of a much broader

1:21:48

set of things that organisms are doing,

1:21:52

exploiting affordances in their

1:21:54

environments. And taking advantage of

1:21:56

opportunities and avoiding threats in

1:21:58

a way. And it's not just that.

1:21:59

just guppies creating dense, nasty

1:22:02

competitive streams, it feels

1:22:04

like this is a characteristic of life everywhere.

1:22:07

And that that agential thinking somehow

1:22:09

sort of unifies those various ways by

1:22:11

which organisms are interacting with their

1:22:13

environments. Yeah, I mean, it becomes very

1:22:16

specific, right? The amazing thing about agency

1:22:18

is that

1:22:19

agency is directed at something, it's for

1:22:21

something. I know to the theology, we're

1:22:23

not supposed to

1:22:25

simplify such things, but it is for

1:22:27

something, which means that the kind of things

1:22:29

that you could expect to find in populations

1:22:31

and subsequently how they're going to involve,

1:22:33

it's not just anything,

1:22:35

right? There's gonna be particular paths that lineages

1:22:37

can take because of the kind of challenges that they have.

1:22:40

Again, this variation, it

1:22:42

doesn't just become any old thing, it becomes that thing,

1:22:44

that physiological problem that

1:22:46

agency is for. Yeah,

1:22:49

and again, I mean, so like,

1:22:50

at least in the literature that I read, I

1:22:53

see people talk about how

1:22:56

contingent is evolutionary

1:22:58

change based on the past. Like

1:23:01

that is a topic that is thought

1:23:03

about, what are the constraints? What are the

1:23:05

trade-offs? How do those dictate

1:23:08

the direction that evolution

1:23:11

will go

1:23:12

go towards? Is there a

1:23:14

bias because of some kind

1:23:16

of historic

1:23:17

sort of constraint like that? And so,

1:23:19

if the concept of agency can be

1:23:22

brought into that

1:23:24

type of thinking and say like, given

1:23:26

what we know about this particular lineage,

1:23:28

for example, and

1:23:30

agency is one of the other

1:23:32

kinds of baggage that they bring with them, sometimes

1:23:36

that baggage can be not so good because

1:23:39

it really prevents you from exploiting

1:23:42

certain kinds of environments and evolving

1:23:44

in certain directions, but then

1:23:47

agency could certainly be

1:23:50

thought of as a way of, for example, facilitating

1:23:54

evolutionary change into, for

1:23:56

example, colonizing new environments because

1:23:59

of some kind of

1:23:59

something that's there, but you mentioned

1:24:02

the word teleology and teleonomy,

1:24:05

like, I think that's where a

1:24:07

lot of people get uncomfortable within

1:24:09

the evolutionary community because because

1:24:12

then it implies that there is

1:24:14

some, you know,

1:24:15

known outcome that

1:24:18

will happen. And, and I think, repeatedly,

1:24:21

the evidence suggests that, you

1:24:23

know, there isn't necessarily going

1:24:25

to be evolutionary progression

1:24:27

towards, for example, greater complexity

1:24:30

or greater agential sort

1:24:33

of behavior. I mean, we, we see,

1:24:35

for example, that there are some lineages

1:24:37

that have more or less been

1:24:40

unchanged for, you know, millions

1:24:42

and millions of years, living

1:24:44

in, in very

1:24:45

constant, usually kinds of environments

1:24:48

with no, no change. And then, you know, we've

1:24:50

seen other lineages that have like,

1:24:53

diversified and, and expanded.

1:24:56

And so, I don't think agency would

1:24:58

necessarily predict progress or complexity.

1:25:00

As long as the system remains viable, it doesn't

1:25:03

need to become anything different. If the world

1:25:05

is predictable, it remains, it maintains

1:25:07

the same old model. And I think that's totally

1:25:09

fine. But if and when the world does change, it

1:25:11

has to act back or it has to update its model. I don't

1:25:14

think that those ideas are incompatible. But bringing

1:25:16

up this idea, it reminds me

1:25:18

to ask you about something that you mentioned a couple

1:25:21

of different times. And I think it's, it's a part

1:25:23

that I feel a little bit responsible to touch. Do

1:25:25

you think that talk or where do you think in 2023,

1:25:27

the field stands with agency

1:25:30

as some kind of spooky force?

1:25:33

Like, do you feel that biologists have to be

1:25:35

extra careful about using words like

1:25:37

this because of the potential influence it has on

1:25:40

folks that are against or have different agendas,

1:25:42

like intelligent design, that kind of thing?

1:25:44

Well, I'm not an expert

1:25:47

on these kinds of ideas. And I'm sure

1:25:49

there are other people who've thought about this a lot more.

1:25:52

I can only say from my perspective, one

1:25:54

thing that makes me very uncomfortable is

1:25:57

that

1:25:58

most of the papers that I've read

1:26:00

that deal with agency are

1:26:02

funded by the Templeton Foundation.

1:26:05

And so within that group,

1:26:07

within that set of researchers and

1:26:10

philosophers

1:26:12

and biologists who are funded through

1:26:14

that agency. So

1:26:21

does that mean that it's just

1:26:23

a little bubble of people that

1:26:26

have

1:26:27

similar thinking and

1:26:29

are all in agreement with one another and

1:26:32

there's nothing more than that? That could be

1:26:34

the case. What makes me

1:26:36

uncomfortable and I think makes a lot of

1:26:38

other people uncomfortable is this

1:26:40

way of depicting evolutionary biology

1:26:43

as being somehow in crisis and

1:26:45

that there's a real problem here

1:26:47

and there's some fundamental

1:26:49

problem that needs to be corrected.

1:26:53

And whether it's what people are

1:26:55

calling the extended evolutionary synthesis

1:26:57

or using new terms like agency.

1:27:01

So within biology, I don't think this is a big

1:27:03

deal but does this open the door for the general

1:27:05

public to then say, look, biologists

1:27:08

evolutionary biology

1:27:10

is all wrong or has been

1:27:12

wrong and is

1:27:15

in this crisis mode and that

1:27:18

somehow

1:27:19

gives the perception that it

1:27:21

gets discredited, it shouldn't

1:27:23

be trusted because those guys still haven't

1:27:25

figured things out. That gives me strength

1:27:28

to push maybe

1:27:30

a creationist or intelligent design

1:27:33

kind of agenda. Those

1:27:35

are our concerns. I mean, those are, there

1:27:37

are people out there who would exploit these kinds

1:27:39

of disagreements in that kind of way. So

1:27:43

I think it's extremely healthy to,

1:27:46

we don't want everybody to have the

1:27:48

exact same views. We want people

1:27:50

to have a diversity of views. Wouldn't be science, wouldn't it? Exactly.

1:27:54

And we want to encourage that. And so, I

1:27:57

think the challenge here is articulating.

1:27:59

and incorporating a view of agency

1:28:03

that is compatible

1:28:05

with and appealing to those

1:28:08

people who are actually doing

1:28:10

evolutionary biology research and

1:28:12

how would you incorporate it into you

1:28:15

know your studying guppies

1:28:18

or insects or birds? Would

1:28:20

it fundamentally change the

1:28:22

kinds of experiments you design,

1:28:24

you know the assumptions that you're making, the interpretation

1:28:27

of your results in ways that the

1:28:29

standard theory wouldn't do? I do

1:28:31

hear your concerns, I mean and it's unfortunate

1:28:33

that you know a lot of the people talking

1:28:36

about agency are also funded by you know

1:28:38

this agency that could raise

1:28:40

some

1:28:41

some doubts. I mean I agree

1:28:43

with you there on camera. Earlier

1:28:45

you said that

1:28:47

a grassland ecosystem could have

1:28:50

agency. It's hard

1:28:52

for me to

1:28:54

see how

1:29:04

you use the term without cognition.

1:29:07

A cognition for sure because any

1:29:09

any complex system with the ability to update

1:29:11

is learning

1:29:12

so that's just that's cognition. But I

1:29:15

think you guys may differ on this meaning of cognition

1:29:17

too right? Oh probably yeah I mean the

1:29:19

last the last part of the conversation we

1:29:21

focused on the thing we just recorded was

1:29:23

with Mike Levin that was it called

1:29:25

cognition all the way down? I thought it was

1:29:28

agency all the way down. Agency all the way

1:29:30

down but cognition is the is the main word that

1:29:32

he uses in there. And the people that are invoking

1:29:34

cognition this way are I

1:29:36

think generally careful while they're rare

1:29:38

probably for a good reason and they're

1:29:40

very careful to to distinguish

1:29:42

it from consciousness right? So there's nothing about

1:29:45

cognition that that requires

1:29:47

consciousness. Awareness is secondary.

1:29:50

Yeah yeah but what so

1:29:52

guys we've been going on for a little while and

1:29:54

I've got to go in a few minutes what are the

1:29:56

like main landing points that that

1:29:58

we want to hit? Cam is there something? Well,

1:30:00

I think one thing that I wanted

1:30:02

to say was that one thing

1:30:04

that I've really appreciated about

1:30:08

big biology as a listener

1:30:10

and now joining as a co-host is

1:30:13

interviewing people with diverse opinions

1:30:16

for sure, even if I don't agree with them,

1:30:18

but also interviewing people

1:30:21

who are also interested in the

1:30:24

philosophy of biology. And

1:30:26

I think most working

1:30:29

biologists don't read

1:30:32

philosophy of the science as much

1:30:34

as they probably should. And I

1:30:36

think that's a

1:30:37

really great thing that big biology

1:30:39

has done to introduce working

1:30:41

biologists to this kind of philosophical approach.

1:30:45

But

1:30:45

having said that, I think,

1:30:48

and I know I've heard you and Art

1:30:50

ask these questions in the past is

1:30:53

it's one thing to talk about philosophical

1:30:56

ideas as they relate

1:30:58

to evolution and agency and everything.

1:31:01

It's another much more difficult

1:31:03

task to then bring those ideas

1:31:06

into sort of the working

1:31:09

nuts and bolts of what people actually

1:31:11

do in terms of their research. And

1:31:14

I know you've asked that question in the past

1:31:16

of some of the guests of like, how do I incorporate

1:31:18

that into my own research? So now you're pinning that on

1:31:21

us, I see. And so

1:31:23

now I would just say that I

1:31:25

look forward to another

1:31:28

hundred episodes engaging

1:31:30

with both of you in conversations

1:31:32

about agency and the

1:31:34

complexities of

1:31:37

life

1:31:37

and how life evolves. Yeah,

1:31:40

there's a lot of room for different ideas

1:31:43

and a lot of work that needs to be done.

1:31:45

That's a great place to wrap it up, I think. Honestly. That

1:31:47

is a good place to wrap it up. Did you see everything that you

1:31:50

needed to, Art? Yeah, I think I'm fine. Good.

1:31:53

That was fun. Together, it was fun.

1:32:03

Thanks for listening, and if you like what you hear, please tell a friend,

1:32:05

mention us on social media, or if you're feeling really

1:32:07

generous, remember we're a non-profit.

1:32:10

We always welcome donations to help support our production

1:32:12

team, and especially our student interns.

1:32:15

And a special thanks to our dedicated fans for helping us reach

1:32:17

this 100th episode. Without your support

1:32:19

and enthusiasm, we'd never have made it here. Thanks

1:32:22

to Steve Lane who manages the website, and Ruth Demry

1:32:24

for producing the episode. Thank you as well to

1:32:26

interns Dana Delec-Cruz and Kyle Smith,

1:32:28

who helped produce the episode. Kating Shamiri does our

1:32:31

awesome cover art.

1:32:32

Thanks to the College of Public Health at the University of

1:32:34

South Florida, the College of Humanities and Sciences

1:32:36

at the University of Montana, and the National

1:32:39

Science Foundation for support.

1:32:42

Music on the episode is from Poddington Bear and Tearan

1:32:44

Castello.