Automated Reasoning, Fast and Slow^*

Natarajan Shankar

Abstract. Psychologists have argued that human behavior is the result of the interaction between two different cognitive modules. System 1 is fast, intuitive, and error-prone, whereas System 2 is slow, logical, and reliable. When it comes to reasoning, the field of automated deduction has focused its attention on the slow System 2 processes. We argue that there is an increasing role for forms of reasoning that are closer to System 1 particularly in tasks that involve uncertainty, partial knowledge, and discrimination. The interaction between these two systems of reasoning is also fertile ground for further exploration. We present some tentative and preliminary speculation on the prospects for automated reasoning in the style of System 1, and the synergy with the more traditional System 2 strand of work. We explore this interaction by focusing on the use of cues in perception, reasoning, and communication.

^*This work was supported by NSF Grant CSR-EHCS(CPS)-0834810, NASA Cooperative Agreement NNA10DE73C and by DARPA under agreement number FA8750-12-C-0284. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of DARPA or the U.S. Government. Sam Owre and Shalini Ghosh provided insightful feedback on earlier drafts of the paper.

LNAI 7898, p. 145 ff.

Full article in PDF | BibTeX

lncs@springer.com
© Springer-Verlag Berlin Heidelberg 2013