Interactive slide selection algorithm and machine learning in psychophysiological memory testing

Area of Science:

• Psychology
• Forensic Science
• Biomedical Engineering

Background:

• Standard concealed information tests (sCIT) present stimuli sequentially.
• Electrodermal activity (EDA) is a key physiological measure in deception detection.
• Optimizing stimulus presentation based on EDA could enhance CIT accuracy.

Purpose of the Study:

• To introduce a novel concealed information test (CIT) utilizing an interactive slide selection (ISS) algorithm.
• To compare the effectiveness of the ISS-based CIT (issCIT) against a standard CIT (sCIT).

Main Methods:

• The ISS algorithm interactively selects slides based on real-time electrodermal activity analysis.
• sCIT uses a predefined, sequential order for slide presentation.
• 64 participants were tested, with 32 in the sCIT group and 32 in the issCIT group, using objects like bags and wallets.

Main Results:

• The ISS algorithm demonstrated significantly better true/false predictions (p<0.01) and fewer false positives (p<0.001) than sCIT.
• Machine learning classifiers boosted precision from 49% to 79% (sCIT) and 85% to 100% (issCIT).
• Testing time was substantially reduced with issCIT (average 53s) compared to sCIT (330s).

Conclusions:

• The ISS algorithm offers superior classification accuracy and efficiency for concealed information testing.
• Machine learning integration further optimizes classification performance, reaching 100% precision.
• ISS-based CIT presents a more effective and time-efficient alternative to standard methods.