In thrall to games?

Hidden Games: The surprising power of game theory to explain irrational human behaviour by Moshe Hoffman and Erez Yoeli does exactly what it says on the cover. It’s an interesting strategy (ahem): irrational behaviour turns out to be rational after all! When being over-optimistic about our own abilities, fooling ourselves by only believing social media comments that reinforce our prior beliefs, spinning the truth, or over-spending on costly luxury goods, all these and more behavioural phenomena can be explained as dominant strategies in appropriately described games. It’s all about the Nash equilibria, stupid.

I enjoyed the book on the whole. It’s a breezy introduction to game theory applications, written with a light touch and plenty of anecdotes. An early chapter sets the scene with the classic evolutionary explanation for observed sex ratios, moving on then to the Hawks and Doves game, and of course the Prisoners’ Dilemma features extensively too.

It has always astonished me how few people think strategically at all, such that a read of the classic (1993) Thinking Strategically on how to apply game theory by Avinash Dixit and Barry Nalebuff would help almost anybody. I can remember once observing in a committee meeting (all economists) what the Nash equilibrium in a certain policy situation involving all EU member countries would be, which did my reputation among those colleagues a power of good as this made it obvious what decision our (UK) politicians would take, regardless of any economic advice we delivered.

Hidden Games does some of the same groundwork as Thinking Strategically in its first half, throwing in a brief explanation of Bayes theorem en passant. I found it less compelling when it gets to the second half, explaining the ‘irrational’ as Nash equilibria in various games. Perhaps this is because there are already pretty powerful models, whether cognitive – eg the rule-of-thumb argument about conserving brain energy in making decisions – or economic – classic signalling models. The application of game theory seems more interesting in contexts of collective choices (as in Kaushik Basu’s wonderful The Republic of Beliefs) than in individual decision-making.

Having said that, if you don’t mind the trope about rationalising the irrational, Hidden Games is a very nice introduction to applying game theory to life, an enjoyable read.


Wolves on the trading floor

My lockdown days have become filled with Zoom meetings at the expense of reading on tubes and trains. I’m definitely going to try to cut down on the online calls, which are exhausting (& I hope tech-land is working out why, and trying to fix it).

Meanwhile, I read a very good biography of Walter Gropius by Fiona MacCarthy, another of the wonderful Maigret novels, and also The Hour Between Dog and Wolf: risk taking, gut feelings, and the biology of boom and bust by John Coates.

It’s a really interesting book about the role biological/neurological responses in decision-making, applied to the financial market context. It will surprise nobody to learn that testosterone is one of the key players, generating bull markets (hah!) and excess risk-taking. “Traders are walking time bombs, and banks invariably light the fuse, dangling before them huge risk limits and bonus payments.” There seem some obvious regulatory interventions in the financial context eg ban bonus structures and mandate 50% female employees on trading floors. The book points out that at most 5% of traders are women, even though they outperform men over the long term.

The book braids together sections on the biology and sections tracking the various hormones and nervous impulses in a financial market boom. It’s a terrific read and super-clear. The author is a financial markets guy turned research scientist and having both sets of insights is illuminating.

The part that most interested me though – and that put me on to the book via an FT article about the impact of uncertainty on our health – was pondering what it means for a computer to think and make decisions when human decision-making is so firmly embodied, driven by our physical features, the way the chemicals in the blood stream and the nervous signals shape perception and emotion. The book I’ve now started (Economic Life in the Real World by Charles Stafford ) cites Antonio Damasio’s work in Descartes’ Error: people whose emotions are affected by brain injury are worse at making ‘rational’ decisions. Reason and emotion – involving biochemical and neurological phenomena – go hand in hand. Meanwhile, we are building AIs according to an idea of ‘reason’ modelled on homo economicus.


The importance of small decisions

The importance of small decisions (by M O’Brien, A Bentley and W Brock) is a neat little primer on an evolutionary approach to decision theory. I read it under the influence of severe jet lag and so nothing at all should be read into the fact that I find it hard to summarise the argument – there’s also an awful lot of explanation by way of analogy to American football, which doesn’t help me at all. My takeaway is the title, plus this very neat diagram putting decision making modes into four quadrants separated along the dimensions transparent/opaque and individual learning/social learning:


It’s a short book so anyone whose interest is piqued will find it easy to read (sporting analogies aside).

[amazon_link asins=’0262039745′ template=’ProductAd’ store=’enlighteconom-21′ marketplace=’UK’ link_id=’1ccd5cd0-b34d-4dc2-9f55-2058015524c3′]

Cognitive engineers, not choice architects

Although I’ve been reading at least as much as ever, it’s been difficult to find time to post about the books, given the length of my To Do list and the depth of untackled emails. One book I finished a while ago and wanted to flag up was Cognitive Gadgets: the cultural evolution of thinking by Cecilia Heyes. This is outside my territory, but as she incorporates cognitive science and linguistics in her work, I’ll be bravely inter-disciplinary too.

The book’s argument – as the subtitle flags up –  is that humans’ distinctive cognitive abilities are due to cultural evolution rather than genetic. It considers the evidence for cognitive differences between human babies and newborn chimps, and concludes that are rather subtle although importantly including a greater human ability to learn, and to remember. Then as humans grow we acquire our far greater distinctive cognitive skills from the society around us – they are not encoded in our minds, they are not simply shaped by social learning, but rather mechanisms or ways of thinking that have been built by cultural evolution: “They are cognitive gadgets rather than cognitive instincts: pieces of mental technology that are not merely tuned but assembled in the course of childhood through social interaction.” Some parts of the gadgetry will of course be inherited genetically, but the assemblage is the result of natural selection operating on cultural variation.

The book argues that this approach overcomes one of the issues with the idea of memes, because the issue is what units are memes measured in? What does the force of natural selection actually operate on – tunes, ideas and other memes? This seems unlikely. Heyes suggests the memes are grist to the mill of cognitive mechanisms such as causal understanding, imitation, reading aloud, causal inference, language, and other ‘gadgets’. The best gadgets thrive in the cultural evolutionary process. When it comes to inheritance mechanisms, it is social and cultural learning. This can largely occur inside individuals’ minds, but can also involve social processes such as story-telling, learning to take turns, group dances, even teaching.

History suggests that these have been quite robust – here we are, after all, with all the warp and weft of modern life. However, the theory does suggest a certain vulnerability, as many of these mechanisms could fail to be passed on: “The cognitive instinct view implies that human nature is relatively invulnerable to catastrophe. In a decimated and isolated human population … the group would lose some of its knowledge and skills. However, with each birth there would be a new child equipped with Big Special cognitive instincts. … In contrast the cognitive gadgets view implies that both grist and mills would be lost.  … The capacity for cultural evolution, as well as the products of cultural evolution, could be lost.” We are already failing to pass on to many children some quite basic (you’d have thought) gadgets such as critical thinking. Heyes specifically suggests looking at how moral learning has evolved and continues to do so.

Anyway, amateur that I am, I found this a persuasive approach and a really interesting book to read. It suggests a different approach to thinking about decision-making – not for example as a matter of setting up choices in ways that nudge flawed humans to do the right thing (but then what makes the choice architects any wiser?). Instead the engineering challenge is devising better gadgets, which is surely difficult but then humanity has invented them before.

[amazon_link asins=’0674980158′ template=’ProductAd’ store=’enlighteconom-21′ marketplace=’UK’ link_id=’e8e8ca64-804c-11e8-8bf3-c5e25527bf61′]



Civilisation: primeval slime to Mars

I finished Daniel Dennett’s From Bacteria to Bach and Back, and personally have no problem with his view that human consciousness is an evolved characteristic built over time from the ground up, and that human culture evolves too, starting with language and continues through memes. In other words, it’s all cranes, not skyhooks. Some people are obviously troubled by this argument. I lack the technical knowledge to evaluate all the detail here. It fundamentally seems far more plausible to me than the alternative.

As a matter of logic, this requires me – and Dennett – to take seriously the argument that computers/AI could evolve minds and consciousness. He puts some weight on the importance of embodiment – but that might be possible although we’re not there yet. Computer vision would be different from ours, but then so is flies’ vision or cephalopods’.

More of an issue, it seems to me, is that computers/AI are very energy-hungry compared to our brains: at the moment, Dennett writes, computer intelligence is parasitical, depending on humans to feed them a lot of energy and otherwise maintain them. What’s more, computers don’t have to struggle or compete: “Down in the hardware, the electric power is doled out evenhandedly and abundantly; no circuit risks starving. At the software level, a benevolent scheduler doles out machine cycles to whatever process has the highest priority, and although there may be a bidding mechanism … this is an orderly queue, not a struggle for life.”

So for now, I’ll stick to thinking Singularity-talk is mystical hype; but will try to keep an open mind on this question.

I’ve always had a soft spot for memes. Dennett uses words as the paradigmitic examples. It reminded me of a jokey line I read once about libraries being the dominant life form on Earth because they are so good at finding new hosts who will start to accumulate books.

There’s a nice section on the importance of social trust at the end of Bacteria/Bach, citing Paul Seabright’s wonderful Company of Strangers. Trust is the invisible glue of human societies, Dennett writes, and much too recent to be a hard-wired natural instinct. “We have bootstrapped ourselves into the heady altitudes of modern civilisation, and our natural emotions and other instinctual responses do not always serve our new circumstances. Civilisation is a work in progress and we abandon our attempt to understand it at our peril.”

Looking at the news these days, that isn’t a very optimistic note on which to end. And yet yesterday brought the amazing launch of Space X’s Falcon Heavy. Astonishing. Perhaps we’ll end up on Mars while the computers colonise Earth.

[amazon_link asins=’014197804X’ template=’ProductAd’ store=’enlighteconom-21′ marketplace=’UK’ link_id=’b39737f3-0c32-11e8-ad96-a76f5f94eca4′]  [amazon_link asins=’0691146462′ template=’ProductAd’ store=’enlighteconom-21′ marketplace=’UK’ link_id=’bd13ce93-0c32-11e8-83e6-1bec8223b6a1′]