Humans as information economisers (I)

Satisficing economisers will develop money, suffer money illusion, and not maximise their expectations.

Hulki Okan Tabak, Unsplash.

The twelfth essay in the Worshipping the Future series is a critique of Economics for being insufficiently congruent with the anthropogenic sciences.

Arnold Kling has provided useful criticism and comment, to which I have responded.

A model used in mainstream economics is of the rational maximising agent with complete information. It is highly mathematically tractable. It makes no biological sense.

Biological beings rarely maximise, for another word for maximisation is exhaustion. To go to the limits of capacity, so to exhaustion, is often to risk death.

Rather, biological beings economise in order to satisfice. To deal with what is currently salient so they can, and have the resources to, move on to the next salient thing.

Biological beings do persistently economise, but they economise to avoid exhaustion, to avoid “maxing out”.

Complete information models also have many problems.

Complete information is typically neither possible nor adaptive if it were. Biological beings economise on everything, including perception, cognitive capacity and cognitive effort. As they economise so they have resources to deal with the next salient thing, over-investing in perception capacity is not conducive to lineage survival. There are reasons lineages in lightless environments can adapt across generations by losing their eyes.

While I am highly critical of the rational maximising agent model, I am also unimpressed with much of the work on human “irrationality”. Any result which implies that human decision-making is not robust enough to survive dangerous environments is evolutionarily implausible.

We have evolutionarily invested to an extraordinary degree in cognitive capacity. To the extent of having unusually dangerous childbirths (due to getting large infant heads through bipedal pelvises), unusually helpless infants (basically, brains with food and elimination mechanisms) and unusually long childhoods with much brain development taking place after birth across two decades.

This investment in cognitive capacity drives the basic structure of human societies of transferring risks away from child-rearing and resources to child-rearing. It must be expected to be worth all that investment. Hence we should be sceptical of any results that imply it is a failed investment.

Even if you want to argue it is flawed, as economising on cognitive effort will be, it is implausible that it is not good enough to cope with the wide range of environments foraging humans can demonstrably flourish in.

Various studies finding such lacking-in-robustness decision-making have failed to replicate. For instance, that hunger affects judges’ sentencing decisions. This never made much evolutionary sense. We are the highest-body-fat share primate precisely so we can go longer periods between eating while sustaining our energy-hog brains (which demand between a fifth and a quarter of our total calorie consumption).

What makes evolutionary sense is us operating a bounded rationality. As us being able to make decisions conducive to personal and lineage survival, given limited information, limited cognitive capacity and limitations on attention. Such an organism will, for instance, develop habits as ways of economising on attention and cognitive effort.

One of the problems with maximising rationality with complete information is: why do we have memory and learning if not to improve decision-making? Maximising rationality is a blank slate conception that tends to devalue experience, and so the child-adult distinction, and not grapple with human variation or with path dependence in the evolution of human cognitive capacity.

As Herbert Simon put it in his 1978 Nobel Memorial lecture:

Human behavior, even rational human behavior, is not to be accounted for by a handful of invariants. It is certainly not to be accounted for by assuming perfect adaptation to the environment. Its basic mechanisms may be relatively simple, and I believe they are, but that simplicity operates in interaction with extremely complex boundary conditions imposed by the environment and by the very facts of human long-term memory and of the capacity of human beings, individually and collectively, to learn.

For example, a parsimonious model of human cognitive biases is that they come from paths of investment in beliefs. That:

several—so far mostly unrelated—biases (e.g., bias blind spot, hostile media bias, egocentric/ethnocentric bias, outcome bias) can be traced back to the combination of a fundamental prior belief and humans’ tendency toward belief-consistent information processing. What varies between different biases is essentially the specific belief that guides information processing. More importantly, we propose that different biases even share the same underlying belief and differ only in the specific outcome of information processing that is assessed (i.e., the dependent variable), thus tapping into different manifestations of the same latent information processing.

In the environments in which we evolved, such information processing would be subject to regular reality-tests. It took decades of training and experience to reach peak foraging capacity. Where reality tests are weaker, more cognitive biases can be expected to manifest.

We are the products of selection for the capacity to make decisions that lead to lineage survival. There is no reason to think that will be to maximise subjective expected utility, even as a proxy. On the contrary, a certain amount of self-sacrifice will be conducive to lineage survival (notably investment in children, in kin, in the group).

Ronald Coase was surely correct to say that utility was merely a placeholder for a more developed theory of human choice. A theory that can only be properly developed from the evolutionary perspective.

We Homo sapiens are the biosphere champions in non-kin cooperation. While social species are fantastically successful, being a remarkably high proportion of animal biomass, they are phylogenetically rare, as it is difficult to shift from lineage competition to systematic cooperation. Indeed, in the case of eusocial insects, they solve the problem by not doing so: the sterile worker and soldier daughters serve the lineage of their queen and mother.

Individual variation is a fundamental feature of all sexually reproducing species. Yet rationality maximising complete information leads naturally to single agent models that eliminate human variation

Human variation is a large part of why search, discovery and learning is so fundamental to so many social processes, including commerce, especially employing people.

As both an employer (in a very small way) and someone paid to present in entertaining and educational ways, human variability is fundamental to employing and managing people and in teaching.

When it comes to transferring resources, the basic forms are coercion, exchange, connection and structured sharing (“pooling”). Gifts, for instance, are investments in connection. Taxes are coerced transfers while building foyers, parks, roads, etc. represent structured sharing.

In any situation where one is interacting over time, there is a connection to be managed. This is a persistent feature of all contracts, but particularly employment contracts.

As it is impossible for a contract to specify responses to all eventualities, employment relationships work if both sides “invest” in the relationship. If both employee and employer do things that are not specifically contractually required, but keep the relationship functional. The more complex the production tasks, so the more is not covered by specific contractual obligations, the more that is so.

It is precisely the need to keep employees investing in the employment connection that help make wages “sticky” downwards. To cut wages is to very ostentatiously reduce one’s investment in the relationship: this is absolutely an invitation for one’s employees to do the same. It is typically much better to simply terminate a minority of the employment connections than to poison all of them.

If an employer is not worried about employees defecting to other employment, whether because they are sufficiently invested in continuing the connection or prospects elsewhere are also murky, then wages can also resist upward pressures.

Coping with change

Resilience is able to persist (though changing circumstances). Efficiency is how much can be produced from how much resources. There is often a trade-off between strengthening capacity to operate across time and how much output is produced now.

Not valuing institutional or corporate memory is classic sacrificing resilience for short-term efficiency. Creating supply chains that prove highly vulnerable to changing circumstances also represent valuing efficiency over resilience.

Moreover, there is a difference between domain-maximising (maximising across a particular range or area of action) and capacity-maximisation (i.e. exhaustion that threatens resilience). An agent may domain-maximise while still capacity-satisficing, thereby achieving domain-maximising efficiency that does not seriously compromise capacity-satisficing resilience.[&]

Fragility is being easily damaged by change, robust is being indifferent to change, anti-fragility is getting better with change. Being anti-fragile is generally a feature of systems rather than individual objects. Something can be fragile, robust or anti-fragile depending on the type and rate of change.

You want systems that incorporate resilient combinations of fragility, robustness and anti-fragility. Mercantile systems tend to be quite resilient, with elements of anti-fragility, while tending to select for efficiency, as releasing unused or underused resources generate incomes. Hence, various trade patterns persisted for centuries, despite periodic interruptions by war or natural disaster.

Command-and-control systems tend to be brittle: apparently robust until they suddenly fail, potentially catastrophically. They also tend to select for inefficiency, as the more resources spent on their administrative processes, the better for the administrators, and be poor at discovery.

Causing opposing systems of command-and-control to fail is a fundamental aim in military conflict. This is a key reason why Western armies tend to dominate conventional battlefields: their soldiers are trained, and their military systems organised, to cope with chaos far better than their opponents. Hence Western soldiers with no experience of conventional war crushing an Iraqi army full of veterans of the Iran-Iraq War (1980-1988) in the 1991 Gulf War.

Similarly, drill made Roman armies, and (after Maurice of Nassau’s reintroduction of drill in the late C16th) European armies so much more effective on battlefields. Drill habituates soldiers to respond to orders, so makes command-and-control much smoother: it hugely lessens friction in responses to orders.

Information levels

Living organisms use information and resources to maintain themselves, and we humans are unusually self-conscious precisely so we can package and receive information. So we can define certainty, risk and uncertainty in terms of information.

Certainty is a proposition or conclusion that cannot be changed by any new information. Moral certainty (aka beyond reasonable doubt) is when such new information is highly unlikely. Risk is incomplete information that is still sufficient to calculate likelihoods. (Though that may rely on the presumption we have correctly picked what distribution of possibilities applies.) Uncertainty is so little information that we cannot calculate likelihoods.

In the absence of the ability to calculate risk, we use heuristics to cope with uncertainty. Uncertainty is only paralysing when we lose confidence in available heuristics. In that case, people retreat to assets about which there is dense and reliable information, valuing that over rates of return.

Expectations and predictions

Suppose prices incorporate all existing information. Then you can accurately predict on their basis only if no new information arises. What you can infer from existing prices is the embedded expectations.

A prediction that fails to incorporate information currently incorporated in prices is even less likely to come true, given that it will almost certainly fail without any new information. Indeed, it may generate expectations likely to sabotage, rather than promote, economic stability.

The notion of complete information rational maximisers takes time (so learning, investment, attention limitations and human variability) out of the analysis. Satisficing incorporates all of that. The rational expectations revolution does not fully grapple with the ways we economise on cognitive effort, with how much we are cognitive misers.

(We haven’t got to money yet: that is in Part 2.)

Footnotes

[&] Added paragraph.