Escalation of Commitment Simulator

This a Bayesian Updating application that simulates the escalation and persistence of commitment in simulated, no-win situations. Simulated decision-making data are produced in the structure of a sequential learning task (i.e., series of Wins, Losses).

This simulated context suggests that escalation of commitment may emerge following individual histories of reinforcement (e.g., Wins) rather than from flawed cognitive processes or heuristics.

Gilroy, S. P., & Hantula, D. A. (2016). Inherently irrational? A computational model of escalation of commitment as Bayesian Updating. Behavioural processes, 127, 43-51.

Escalation Simulator

Escalation as Bayesian Updating

This tool provides the ability to simulate a series of wins and losses. Data are simulated and then Bayesian Updating is applied to lagged learning sequences. Two-, three-, and four-lagged projections are supplied and visualized.

Simulated data takes the form of a sequential learning task. In this randomized sequence of events, a distinct pattern of Wins or Losses take place. From these data, Bayesian Updating is performed in order to determine warranted beliefs.

Warranted beliefs, in Bayesian thinking, refer to the likelihood that an action is likely to succeed taking into account earlier, similar circumstances. As such, warranted beliefs are differentially affected by the amount of exposures as well as the outcomes from each exposure.

To simulate Two-, Three-, and Four-step Markov Models, you may click the Simulate button below. This may take several seconds to complete, depending on the server load.

This simulator is developed using several open-source components. These tools are listed below along with their respective licenses. All core methods are also available on Github.

ISAAC, public domain (PHP-port, Illmari Karonen; OG, Bob Jenkins)

Chart.js - MIT Licensed

Jquery, MIT Licensed