Basic HTML Version

Emotional Coherence
The standard view of inference from philosophy and artificial intelligence is sequential: you
start with some beliefs, you apply rules and inference, and you acquire more beliefs. The
coherence approach to this is different. It is inspired by neural network models and it says
that what happens when people reach conclusions is in fact much more holistic, but not in a
mystical way. You have a collection of different beliefs that are connected to each other in
different ways and you come up with a coherent account of how they should operate. So
coherence is essentially the normative theory of how you ought to make inferences.
Of course, people do not always behave normatively; sometimes they do things that
are not sensible. Every element in our network of coherent or incoherent elements is going
to have not only a degree of acceptance, which is analogous to a probability, but also an
emotional valence, which could be thought of as desirability. When you are thinking about
something like whether stocks are going to keep going up, it is not merely that you think it
is probable or improbable. You are also thinking about whether it is desirable or
undesirable, and that is factored into the conclusion. When people make decisions, they are
doing it on the basis of emotional coherence and they need to have some way of getting
utility into the picture, not only probabilities. How are you going to calculate utility?
That is really hard, because you often cannot be very precise and so you have to go
with your emotional judgments. What you obtain is something like an emotional gestalt, an
overall picture taking into account the emotions of what you should do. Once you have that
picture, you can give an account of how inferences go well, but you can also use it to figure
out the reasons for them sometimes going badly. I have developed accounts of different
kinds of emotional mechanisms (Thagard, 2006) that can get in the way of thinking well.
Any important decision is going to be emotional and often we do it well, but
sometimes we really get it wrong. In economics, you buy and sell stocks based on your
goals, but you can get misled by different kinds of inferences. The one that is most familiar,
because it has been discussed by psychologists for 20 years, is motivated inference. This is
where you let your desires get in the way of your beliefs. Another error pattern is fear-
driven inference. In this case you do not believe something because it makes you feel good,
as in motivated inference, you believe something because it makes you feel bad. The last
error is rage-driven inference, where the actions that you take come about because you are
really angry about something. This governs many political movements, but it is not so
relevant to the economic case.
Motivated inference
Motivated inference is the idea that when you form your beliefs you do it not only on the
basis of the evidence, but also on the basis of goals and desires that distort that evidence. In
philosophy, this is sometimes called wishful thinking, though it is psychologically more
complicated; psychologists talk about positive illusions or the optimism bias, all ways in
which your beliefs are affected not only by the evidence coming in, but by what you want to