Motivational Computing: Transformer-Based Automation of Implicit Motive Coding.

Area of Science:

• Psychology
• Computational Linguistics
• Artificial Intelligence

Background:

• Implicit motives correlate with behavior but are difficult to measure.
• Previous automation attempts using marker-word approaches were limited.
• Machine learning offers a promising avenue for automating motive coding.

Purpose of the Study:

• To develop and validate a novel transformer model for automated implicit motive coding.
• To assess the model's performance against established benchmarks and other transformer models.
• To provide a freely accessible API for researchers to utilize the model.

Main Methods:

• A new transformer model was developed for automated motive scoring.
• The model was trained and evaluated on the Winter (1994) training manual benchmark dataset.
• Validation included convergent, divergent, causal, and criterion assessments.

Main Results:

• The transformer model achieved over or near .85 correlation with hand-coded scores (Pearson and intraclass coefficients).
• Evidence of convergent, divergent, causal, and criterion validity was established for machine-predicted scores.
• The model's performance was compared favorably against two other recent transformer-based models.

Conclusions:

• The developed transformer model offers a highly accurate and validated method for automating implicit motive coding.
• This work contributes to the emerging field of
• motivational computing,
• facilitating real-time motive analysis.
• The freely available API will accelerate research in implicit motives and computational psychology.