Cognitive Psychology

TRACS is currently being used as a "mind game" to perform experiments on human thinking at the MITRE Corporation. The research project, called "Mental Models in Naturalistic Decision Making", is investigating how people make diagnoses and decisions in uncertain and time pressured situations. The object is to better understand the fundamental strengths and weaknesses of human thinking, in order to improve the design of computer systems used by military commanders, intelligence analysts and other high stakes decision makers.

So far TRACS research has focused on the simplest game (played solitaire) called "Straight TRACS", and a variation of this game called "Spy TRACS". Future research will focus on "Poker TRACS".

Straight TRACS is a "memory" game. Research on this game has shown that people are remarkably limited in the task of counting cards and "updating odds" - at least when they have to keep track of several different card types (6 types in TRACS) at once. On the left is a plot that shows the results of one experiment. The "actual odds" (for one card type) during the game are shown by the dotted line. The "players' odds", reported by subjects during the game, are shown by the solid black line. The plot shows that players tend to be "anchored" to the "baseline" (initial) odds and that they make relatively minor adjustments to these odds - compared to the adjustments that they should make (shown by the dotted line) as the game goes on (moving to the right on the plot). The solid gray line shows the results of a computer model that was developed to explain and predict this baseline bias of the human mind. The computer model (solid gray line) is in good agreement with the human data (solid black line). For further details click here.

Spy TRACS is a "data fusion" game. The game is like Straight TRACS except that players are given the "deck odds" at each turn so they do not have to count cards. However, players are also given odds from a second source (the spy) and they must combine the "deck odds" with the "spy odds" to get the "final odds" (so they can then choose a card with the best odds). Research on this game has shown that people adopt various "conservative" (non-Bayesian) strategies that prevent them from extracting all the certainty that is available in the data. In life, these strategies can be wasteful in leading people to seek more data than they need to reach a given level of confidence. This cognitive conservatism is of special concern in "high stakes" domains (like military and medical) where data can be expensive and dangerous to collect. Based on insights from TRACS experiments, a visual display called "Bayesian Boxes" (pictured on the left) was designed to help people extract all the certainty that they can from the data that they get when they are faced with "data fusion" problems - like military intelligence and medical diagnosis. For further details click here.

Poker TRACS is a "betting and bluffing game" that simulates the cognitive challenges of business, warfare and virtually every other domain of human endeavor (see John McDonald, "Strategy in Poker, Business and War", Norton, 1950). Poker was the game that launched the whole field of game theory in the mid-1900s. Today Poker is still recognized by experts as the most challenging card game (see John Scarne, "Scarne's Guide to Modern Poker", Simon & Schuster, 1980) and the most similar to real-life (see James McManus, "Positively Fifth Street", Farrar, Straus, Giroux, 2003). Poker TRACS offers additional advantages over standard Poker, and that is why it will be a workhorse for future research on TRACS. For further details click here.

Cognitive-scientific research using card games is uncommon but not without precedent. For example, Robert Abbott writes (in “Abbott’s New Card Games”, Funk and Wagnalls, 1968; also see Martin Gardner’s column in Scientific American, June 1959):

“[The new card game] Eleusis … has created … interest in the scientific world, not only as an enjoyable pastime but also in a wide variety of other applications. For instance, a research psychologist at one of the large defense systems corporations is using Eleusis as an experimental medium with which to study inductive reasoning in humans. He wrote me that one of the reasons for his interest in Eleusis is that [the card game] is ‘analogous to certain types of military problems that we are vitally concerned with’. One of the things he is investigating is the influence of various machine and display techniques in helping humans to play Eleusis.”

Computer versions of games are particularly useful in this regard, as noted by Dietrich Dörner in his book, “The Logic of Failure: Recognizing and Avoiding Error in Complex Situations”, Perseus Books, 1996:

“Computer simulations [of games] enable us to observe and record the background of planning, decision making and evaluation processes that are usually hidden. It is easier to isolate the psychological determinants of such processes this way that it is to investigate them retrospectively in the real world. We … find that the sources of [human] failings are often quite simple and can be eliminated … having identified and understood these tendencies … [in games].”