Skip to ContentSkip to Navigation
About usNews and EventsEvents and open daysPhD ceremonies

PhD ceremony Mr. E. Nyamsuren: From vision to reasoning. What it takes to model a computer game opponent

When:Fr 28-03-2014 at 14:30
Where:Academiegebouw, Broerstraat 5, Groningen

PhD ceremony: Mr. E. Nyamsuren

Dissertation: From vision to reasoning. What it takes to model a computer game opponent

Promotor(s): prof. N.A. Taatgen

Faculty: Mathematics and Natural Sciences

I conducted an opening study of types of decision and reasoning processes involved while people play a computerized version of the board game called SET. An extensive analysis of participants' eye movement data and of other behavioral measures revealed that visual perception can directly influence decision making process. However, the study also revealed that current cognitive architectures, ACT-R among them, are not sophisticated enough to model complex tasks such as SET. The study clearly indicated that two major components were lacking in ACT-R.

Following these results, more specific goals were (1) proposing a unifying theoretical framework for top-down control (decision-making and planning) and bottom-up perceptual processes (visual attention control) and (2) developing software tools capable of simulating above-mentioned processes in a way independent of any specific task paradigm. I conducted behavioral experiments aimed at investigating processes involved in visual attentional guidance during a problem-solving task. I have developed two modeling tools that incorporate theoretical findings from my research: the Pre-Attentive and Attentive Vision (PAAV) module and the Human Reasoning Module (HRM). PAAV is an implementation of human vision theories arguing for sequential attentive visual processes operating on top of parallel pre-attentive processes. The HRM proposes a unifying framework of human reasoning. It argues that various facets of human reasoning can be explained via deterministic inference on stochastic knowledge. The PAAV and HRM are part of ACT-R cognitive architecture and can be used to model human behavior in a wide range of tasks.

printView this page in: Nederlands