11 Comments
Dec 26, 2022Liked by Arnold Kling

Arnold;

My subject of professional study is how biology learns - which is useful because mathematical abstractions too often founder on the 'spherical chicken' and sociology too often founders on mood affiliation.

If you pardon some grammatical cleverness, I'd like to address the question of what 'showing your work' looks like in a world where very very few people or organisms are professional thinkers. Your notion of figuring out who to believe based on them writing out their explicit logic is sound in a scholastic or scholarly environment, but most organisms or people show their work by literally practicing it. If it works for them in the short run, you can observe that. If it works for them in the long run, you can observe that - more slowly. If it works in the very long run, you can observe their progeny.

Girard is indeed an astute master and he observed that this all works very well for certain key kinds of traits. Essential behaviors such as avoiding poisonous food are well learned by observation. Lineages that teach and thus preserve them are successful. Of course, it ends up creating conflict over the pre-approved foodstuffs.

However, if there is essential diversity - we might call it 'specialization' - it becomes a downright burden as many people may learn things that are actually ill-suited to them. Further, if some traits are 'green beard' traits - about social identity and affiliation - and each group requires certain social subspecialties at certain frequencies - now we are into economics (or real biology).

The question of 'who to trust' becomes conflated with emulation, learning, and competition. For example, mate competition only makes sense within a single generation, and Freud spent an (inordinate, I think) amount of time contemplating when that behavior is misapplied.

The phrase 'what is good for me' can be ambiguous. We might choose to learn behaviors from someone who exhibits behavior that we predict will work well for us, but that requires an uncommon degree of forecasting and imagination... and draws us away from the strength of the method, that we can simply observe that a successful person does X, and so X must not, at minimum, be immediately fatal to ambition. 'What is good for me' can also be at the level of the society - we can observe that person X produces outcomes which favor us, personally and directly (or in the inverse, damage us, hurt us, cause us pain) and decide to trust or not trust that person accordingly.

Many midwits deride this strategy for learning, saying that it is not intellectually rigorous - but then hypocritically turn around and espouse it when it suits them. In fact, it makes logical sense to trust people who show you sustained respect and effective care; and to emulate them preferentially and favor them for positions of influence.

Anyway, I haven't nearly exhausted the topic, but I hope I haven't taxed your attention.

Expand full comment

This is the best post so far on what you're describing as 'social learning'.

Expand full comment

Just catching up on holiday reads. This is one of my favourite summary posts - I read it twice before buying Secrets for a friend, reading the MH paper, and getting a copy of Laland's book. Thanks very much.

(I realize that this doesn't add a ton to the discussion, but strong appreciation with worth saying. Again, thanks.)

Expand full comment

Social learning doesn't seem so much who to believe, but to observe and see who is being successful and with what, then to copy success steps. It seems like an excellent extension of the Axelrod evolution of cooperation strategy testing - but with rules rewarding copying success more than innovating a possible new success.

I'm pretty sure the idea that too much learning is inefficient is true - but for we who like to learn, it's purpose is not merely to be efficient, but to provide some enjoyment.

From the paper: "Even though INNOVATE cost no more than OBSERVE, the best strategies relied almost entirely on social learning, that is, when learning they almost exclusively chose OBSERVE rather than INNOVATE. "

I strongly suspect the payout function was less realistic than the US market in tech for the last 50 years.

"Under any random payoff distribution, if one observes an agent using the best of several behaviors that it knows about, then the expected payoff of this behavior is much higher than the average payoff of all behaviors, which is the expected return for innovating." To me, this means the Venture Capital dream of a Home Run out of the park never happens due to the structure.

I would like to see how big the top 5% payouts (1 in 20) has to be before an INNOVATE strategy gets lucky and wins big. Instead they adjusted probability (fig. 5):

"payoffs needing to change with 50% probability per round before the INNOVATE-only version could gain a foothold."

At Apple, Lisa failed (too expensive), Mac almost failed, Apple printer big success, Newton failed, iPad & especially iPhone big success. Some strike-outs & some home runs, with more Innovation than most. (Xerox PARC did far too little to Exploit its research). Great link.

Expand full comment

This is also the thesis of Matt Ridley's book "The Evolution of Everything, How New Ideas Emerge"

Expand full comment
founding

Luke Rendell and co-authors write:

"The proportion of moves that involved learning of any kind had a negative effect, indicating that it was detrimental to invest too much time in learning, since payoffs came only through EXPLOIT. The data reveal a surprisingly low optimum proportion of time spent learning [... .]

[...] strategies using a mixture of social and asocial learning are vulnerable to invasion by those using social learning alone [... .]

[...] the strategy that eventually dominates will be the one that can persist with the lowest frequency of asocial learning [...]

[...] our tournament challenges existing theory, which predicts that evolution will inevitably lead to a stable equilibrium where both social and asocial learning persist in a population [... .]

The tournament also draws attention to the significance of social learning errors as a source of adaptive behavioral diversity [... .]"

I have zero expertise in this line of research. I am mindful of the need for cognitive humility (cf. Russ Roberts). Nonetheless, Rendell et al.'s thesis -- observation + imitation + random error suffice to explain human prowess in learning and innovation -- strains credulity.

Two crucial elements are missing from the mix:

(a) Science (an institution)

(b) Entrepreneurship (a kind of creativity, which involves deliberate new recombination/experimentation).

No doubt, my interpretation is dull. A rebuttal might point out that science and entrepreneurship are not instances of "asocial learning." However, if I understand correctly, Rendell et al. define "social learning" in a way that would not include science and entrepreneurship, insofar as science and entrepreneurship go beyond observation + imitation + random error.

Can Rendell et al.'s model explain Paul McCartney, "Yesterday;" the moonshot, or the internet?

Expand full comment