A sideways look at economics
Last week marked my fifth Fathom anniversary, a milestone that passed largely unnoticed in a very busy end of year. Over this period my Fathom inbox has accumulated 14,337 emails, of which 4185 seem to have remained unread (I do receive a lot of junk). Over the past five years, I’ve received emails from 507 different people, while I’ve sent emails to 169 separate ones. I’ve received most correspondence from our chief economist Mr B (1084 emails), followed by administrator Aless (849), and then colleagues Brian, Erik and Kevin with around 300 emails each. What can an external observer infer from these statistics? Probably only that I’m a hoarder, and that Mr B has either got beef with me or likes me (or at times both).
These are objective facts, but they don’t really say much about me or suggest anything substantial about whether I’ve enjoyed these past five years. It is also questionable whether having more data or doing more analysis on this dataset would be helpful in shedding more light on the topic — it might even be counterproductive. Indeed, as economists, we know that the same data can serve multiple masters with different motives. To haters, I often mention that, if nothing else, economists would be useful for unearthing the cheap tricks of those who use data to coat their arguments in a thin veneer of objectivity. There is no Hippocratic Oath in economics, but I feel there should be one, and that one line should read something like: ‘I shall use data and models to always seek the truth.’
Truth-seeking is the objective of any science, but in economics it is hard to separate the seeker from the truth, as both irreducibly deal with human nature. In my view, the subjectivity and uncertainty inherent in social sciences is something that ought to be celebrated. Unfortunately, somewhere along the way, social scientists seem to have become almost ashamed of the ‘social’, in a perhaps too obsessive pursuit of the ‘science’. The cost of doing so has arguably been a drop in relevance, with economists increasingly resembling their stereotype in the ‘assume a can opener’ joke.
Social sciences have gone through existential crises before, and the inferiority complex and reverential respect felt by economists towards the natural sciences have deep roots. For example, in doing research for this blog, I unearthed a fantastic paper from 1940 by Frank Knight (of ‘Knightian uncertainty’ fame) entitled “What is truth” in economics? It was originally written as a scathing review of a book discussing the scientific method as applied to economic theory. It then got published in an economics journal on the merit of its extremely cogent exposition of how social sciences are not only distinctively unique, but why they are perhaps also superior to natural sciences.
Knight subordinates any theoretical notions of ‘objectivity’ and ‘measurability’ to the individuals and their interactions. He even goes so far as to state that anything deemed ‘universal’ can only be so because of a fundamental lack of imagination. For Knight, “we cannot separate the discussion of reality from the discussion of the knowledge of reality”, where, in turn, “all knowledge of the world [is…] a social activity and a social phenomenon”. He goes on that “all such knowledge is inseparable from a) self-knowledge of the knower and b) knowledge of the knowers”.
The idea that knowledge and truth come at least partly from within appeals strongly to me. I have a habit of turning random thoughts into deep rabbit holes that branch into intricate warrens. I enjoy the process of digging such holes as much as extricating myself from them. Truth be told, I’ve often used these TFiF blogs as cheap therapy sessions, to tidy my thoughts about the current set of ideas floating around (more like fighting) in my head. According to Knight, the process of writing these blogs would itself constitute knowledge. They allow the writer to learn about themself, but also to learn from others, and for others to learn from them. Blogging is part of a process where the emergence of “a conscious, critical, social consensus is of the essence of the idea of objectivity or truth”. However, Knight also makes an important qualification about the social consensus: “It rests upon value judgment as to both the competence and the moral reliability of observers. […] If ordinary, normal human beings habitually and systematically lied, or talked dream talk (or reported free association), there would be no possibility of any knowledge.”
I would have been interested to see what Knight made of the emergence of AI. Would he have accepted that AI could advance knowledge by independently complementing the existing social consensus? I don’t think so. I think he would have treated AI as a higher layer of abstract thinking, still conceived by individuals in their pursuit of truth and knowledge. Ultimately if humans can codify human conduct through some algorithm, with the goal of predicting it or shaping it, then humans can also codify and confound AI’s conduct through some other algorithm.
To test this, I’ve asked the openAI algorithm to write two short blogposts: one about whether it would be preferable to lie to oneself or to others, and one about truth in economics. The results are quite impressive, but they lack originality or relatability.
These examples appear more like approximations of the current social consensus, rather than any original elaboration of it. The threat is that in future AI might help to establish and shape the consensus, before one is determined socially. However, tools already exist to counteract such subtle influence, such as the openAI output detector, which in this case was able to spot both blogs as 99.98% likely to have been generated by AI. (This is even more impressive considering that the blogs were written using GPT-3 technology, while the tool was built only to detect its predecessor, GPT-2.) Just as importantly, the tool puts the probability of this blog being written by a human at also 99.98% using only its first 200 words.
Overall, I find it ironic that science is encroaching on the social dimension, after economics has tried so hard for so long to deemphasise it. Looking forward, I’m a cynical optimist: I think AI is a massive opportunity more than it is a threat, not only for society, but also for economics. I am hopeful that we’ll gradually witness a more efficient ‘production function’ in truth-seeking, where science and social science specialise in their respective comparative advantages, and recognise the benefits of working together and complementing one another rather than competing. The threat and the challenge laid out by AI are precisely what ought to make social science and economics intrinsically more valuable to society. Economists who have built a reputation and a credibility by placing the ‘social’ at the heart of social science in their truth-seeking efforts, should be particularly well placed. Look no further than Fathom — and I certainly look forward to the next five years working here.
 In Economics as a Science (1970), Kenneth E. Boulding writes: “There is a story that has been going around about a physicist, a chemist, and an economist who were stranded on a desert island with no implements and a can of food. The physicist and the chemist each devised an ingenious mechanism for getting the can open; the economist merely said, ‘Assume we have a can opener’!”
More by this author: