The Context-Dependency of Human (Economic) Behaviour
Let me start this post by making clear what everybody knows: people behave differently in different kinds of situation, but we can effectively recognise these kinds of situation and use them to understand, and even predict, how to behave and what people will do in these situations. For example we all recognise a lecture and know the social norms, habits, conventions, roles etc. that pertain there. If the lecture is declared finished and coffee or wine served to celebrate something the context has changed and everybody will behave differently.
Perhaps it is the desire to be like physics with universal laws, but economics (like other quantitative social sciences) ignores the context-dependency of human behaviour and cognition. Maybe the fact that people do not only use different information to make decisions in different contexts but seem to decide in different ways, is just too complicated and awkward to acknowledge. It certainly makes understanding and modelling economic behaviour much more difficult!
A classic example of this is how traders behave in bull and bear markets – in the former case they might be seeking to optimise their bonuses (make lots of transactions) and promotion prospects and in the latter trying to keep clients (hence please them) and not be sacked (hence avoid being the worst performer). This is not a question of just using different information to make decisions, but having different goals and making decisions using different algorithms.
More than this, there is ample evidence that other aspects of human cognition are also context dependent, including: memory, language, preferences and perception.
Typically many economic models will try to deal with unmodelled variation in behaviour as if this is some kind of random ‘noise’. However, this is clearly unsatisfactory for many purposes. For example, random noise and context-dependent variation will scale very differently with the size of the system.
The context-dependency of human cognition to be a combination of (A) a rich but ‘fuzzy’ recognition of kinds of situation (B) preferential access to relevant knowledge given this (C) foreground explicit reasoning using that knowledge. One of the problems of this is that (A) is done largely unconsciously and with great fluidity so we are not aware of this (except if we get it wrong).
This context-sensitivity is a very powerful social mechanism in a number of ways, including: (1) kinds of situation can be co-recognised and so people will have access to shared knowledge there (e.g. a lecture) – the more that kind of situation is distinctly recognised the more context-specific technology, norms, expectations, language, habits etc. are developed for it, and so the more recognisable it becomes. This can result in a context becoming socially entrenched. (2) since recognition of context is done on cognitive as well as external queues the processes by which social coordination is achieved can change flexibly with event (e.g. fishing in calm waters and dealing with a storm in the same waters).
The reason that context-dependency has been ignored is its difficulty. Formal modellers might desire that their models are simple and have wide generality – having to included a variety of fundamentally different behaviours for different kinds of situation just makes things hard and messy (and thus unlike those in the physics). Those at the other end – using more qualitative, knowledge-rich approaches, may tend to deal with each case separately – essentially always working within a situation and resisting generalisation – thus avoiding dealing with context-dependency in another way.
However, advances in modelling, big data collection and analysis, and data integration now allows us to start dealing with this ‘inconvenient truth. Ignoring it would be a similar mistake to that of assuming people were rational in the narrow economic sense.