Thinking about chess: An ethological a approach to decision-making

ETCHEMENDY, PABLO; FERNÁNDEZ SLEZAK, DIEGO; SIGMAN, MARIANO

Cordoba

Taller; Reunión Conjunta de Neurociencias (IRCN); 2009

Decision-making has been widely studied in the laboratory. To assure sound statistics the number of choices in these experiments is small. However, in most real-life situations –when buying a pair of shoes, or choosing a vacation destiny - deciders choose within a broad number of options accessing only partial information. Studying decision-making in such ethological context is challenging since it is difficult to evaluate – from a normative or subjective perspective – the elements and the goodness of a choice. Here we capitalized on one of the largest and unexplored databases of complex (many outcomes, many dimensions involved in each outcome) human decision making: chess. Before making a move, a chess player weighs many alternatives in function of its convenience, risk, complexity... leading to a choice. In time-controlled games, the total time is finite and subjects play according to an implicitly chosen time investment policy. A principal advantage of this experimental setup is that the outcome of the decision can be accurately evaluated by contemporary chess software. We have so far downloaded about 500.000 games which roughly correspond to 40.000.000 decisions. This number increases at about 20.000 games per day. We observed 1) a clear pattern of time-investment heavily weighted in the mid-game which indicates strategic fluctuations in the decision criterion 2) a very weak correlation between invested time and outcome of the choice confirming that hunches are at least almost as effective as calculated choices. We present a large number of elements of decision-making which can be explored with this rich database.