Level-K Classification from EEG Signals Using Transfer Learning.

Area of Science:

• Neuroscience
• Game Theory
• Cognitive Science

Background:

• Tacit coordination games involve players making decisions without communication.
• The level-k model describes reasoning depths, where level 0 is random choice and higher levels involve strategic beliefs.
• Focal points are salient solutions that guide choices in these games.

Purpose of the Study:

• To investigate the neural correlates of different reasoning levels (level-k) in tacit coordination games.
• To differentiate brain activity associated with random choice, strategic reasoning, and coordination.

Main Methods:

• A combined behavioral-electrophysiological study was conducted with three conditions: resting state, picking (random choice), and coordination (strategic reasoning).
• Deep learning and transfer learning techniques were employed to analyze the electrophysiological data.
• Machine learning models were trained to classify different reasoning states.

Main Results:

• High precision (99.49%) was achieved in classifying the resting-state condition.
• Classification precision for the picking and coordination conditions was 69.53% and 72.44%, respectively.
• Distinct neural patterns were identified corresponding to different levels of reasoning in tacit coordination games.

Conclusions:

• The study successfully identified neural correlates for varying reasoning depths in tacit coordination games.
• Findings demonstrate the potential of machine learning in analyzing cognitive processes during strategic decision-making.
• Future research can explore applications of these findings in understanding complex human interactions and developing AI.