Evolution of Investment Strategies, and Case-Based Decision Theory

Résumé du document

Gilboa and Schmeidler's work demonstrates an important distinction between behavior as predicted by Expected Utility Theory and as predicted by Case Based Decision Theory: in the event of a posed similar, familiar decision, people will revert to their memories of past situations to determine their current course of action. The implications of their distinctions are interesting, since it suggests the presence of an inclination towards simplifying decisions. Whenever possible, we will attempt to reduce our behavioral choice options by comparing them to choices we have already made and this to me seems like a process designed to both save some cognitive energy and adopt what we have learned from the past.

Extraits

[...] Evolution of Investment Strategies, and Case-Based Decision Theory Gilboa and Schmeidler's work demonstrates an important distinction between behaviors as predicted by Expected Utility Theory and as predicted by Case Based Decision Theory: in the event of a posed similar, familiar decision, people will revert to their memories of past situations to better determine their current course of action. The implications of their distinction are interesting, because it suggests the presence an inclination towards simplifying decisions: whenever possible, we will attempt to reduce our behavioral choice options by comparing them to choices we have already made, and this to me seems like a process designed to both save some cognitive energy and employ what we have learned from the past. [...]

[...] This raises an interesting new element: if we were able to create computer programs in other competitive style games and provide it with a history of our past decisions, it would be able to learn more quickly than us from our past and therefore exploit the weaknesses of our mistakes. In this way, we could potentially use computers to help ourselves determine which past decisions we've been consistently making that are not necessarily the best. This discussion also touches on the theory of satisfying, which states that people will generally choose the decision that satisfies some minimum level of utility for them, without necessarily maximizing utility (i.e. we will make the “first-best” choice that comes our way, possibly because waiting for the best choice may take too long). [...]

[...] In one of the more recent rounds of chess matches between world champion Gary Kasparov and IBM's supercomputer named “Deep a controversy was caused by the way that Deep Blue had been designed: it had been provided with a comprehensive history of every single match that Kasparov had ever played. Thus, whenever presented with a situation that was familiar to a game Kasparov had already played, the computer knew how and where to exploit his weaknesses assuming that Kasparov would play similar to the style he had played in the past. Deep Blue went undefeated in this round of games, however after a prompt set of complaints from Kasparov's team; Deep Blue's programmers were required to shut off the computer's history log. [...]

[...] However, one question that comes to mind is: what is the difference between being able to account for someone else's historical decisions versus being able to account for one's own historical decisions. In other words, Deep Blue exploited the use of someone else's predicted behavior, not the predicted behavior of itself. Kasparov, on other hand, was influenced (probably unknowingly) by his past decisions because the fact that the computer consistently beat him meant that he continued to make the errors he had made in the past. [...]

[...] This theory, coupled with case based decision theory, suggests that if we consistently satisfice, we may be inhibited from reaching maximum utility. This is what my previous argument is premised on, and essentially it seems likely that computers could be used to help us determine when we are “satisficing” versus when we are “maximizing” our utility. While Kasparov may not have liked the publicity of Deep Blue defeating him, he probably would have become a better chess player if he continued to play the original program and therefore learn when he had been incorrectly satisficing. [...]