11 Jan 2018
Whenever I get a really exciting idea, my heart races. I think the excitement and heart-racing is fueled by an idea that seems to vast, so deep, so intractable, that I get lost in my head-space, my heart begins racing, and I search for an escape.
Here’s an example.
Premise 1) Cognitive computational behavioral psychology - the kind of stuff that I did, but barely scratched the surface of during my master’s degree - is so young, and the ideas are so new, that not so many people, organizations, institutions, or businesses have been able to apply the theories of the field.
Premise 2) The Drift Diffusion Model has been applied to response time distributions for two-alternative forced choice tasks, and its been shown that the model is a fairly accurate predictor of behavioral responses in these experiments. To achieve success, the model uses only four parameters (and their deviations): 1) a decision criterion, 2) a mean drift rate for evidence accumulation, 3) non-decision time, and 4) bias.
However, the deadlines for the decision deadlines in the experiments that provide input to the DDM are usually confined to the seconds range. So participants must issue a response within, say, a 4s window. But, what about really important, long-term decisions like buying a car from a choice of 10 cars, or buying stock from a selection of 1000 stock? Or making a move in a game of Go? Is there a model that can predict the variations among people for these long-term decisions?
In other words, given large amounts of time, access to huge amounts of information (e.g. internet), what parameters would a model need to account for the variation of human behavioral responses in real-life, complex situations? This question gets my heart racing.
Let’s call the new model, the Seldon Model (after Hari Seldon, the fictional character from the Issac Asimov Foundation stories). Intuitively, the Seldon Model might include things like sleep quality, food intake, exercise, relationship quality, and many more. These things all seem to play a role in determining the outcomes of our decisions. Since people differ in these variables, though, and since it’s really hard to track these sorts of things, it’s unlikely that any single scientist, organization, institution, or business could do more than speculate about which extra parameters would be needed to inform the Seldon Model.
What the Seldon Model needs is a small amount of efficient parameters that do reasonably well, if not perfect, at predicting the variations of human behavior in real-life, complex situations. A model like this would be very useful, very interesting, cutting-edge, and perhaps even damn-right dangerous. Hari Seldon would admire a person with that sort of model.
What are the chances we develop a Seldon Model?