Scientific Foundation

Scientific Foundations of Chatbot-Assisted Evaluation Tasks

1. Idea Generation Task

This task is guided by the Geneplore framework (Finke et al., 1992), a cognitive model of creativity that distinguishes between two core processes:

  • Generative processes: Creating mental representations or “preinventive structures.”
  • Exploratory processes: Evaluating, transforming, and applying those representations in meaningful ways (Moreau & Dahl, 2005).

The chatbot evaluates the user across four dimensions that align with this generative–exploratory cycle, including collaboration style, critical thinking, novelty-feasibility balance, and decision ownership.

2. Problem-Finding Task

Based on the Augmented Problem-Finding Framework (Sabbah & Li, 2025), this task emphasises:

  • Problem Discovery: Trend analysis, hypothesis generation, and weak signal detection.
  • Problem Definition: Reframing problems, testing assumptions, and refining scopes.

ChatGPT is instructed to assess framework adherence and the depth of human–AI collaboration.

3. Tactical and Strategic Decision-Making Tasks

Adapted from Kline (1996), these tasks involve ranking decision actions. Deviation from expert rankings determines performance (lower score = better). The model analyses:

  • Tactical reasoning patterns
  • Confidence calibration
  • AI reliance
  • Suggestions for future improvement

4. Communication Skills Task

This task employs a 7-item scale adapted from de Vries et al. (2009), known as the PRESENT model:

  • Preciseness: Clarity, conciseness, grammar
  • Reflectiveness: Thoughtful, analytical expression
  • Expressiveness: Eloquence and confidence
  • Supportiveness: Relational and encouraging tone
  • Emotionality: Emotional expressiveness or tension
  • Niceness: Cheerfulness and friendliness
  • Threateningness: Intimidation or manipulation (reverse scored)

5. Negotiation Task

Drawing from Ganesan (1993), this task focuses on four evaluative dimensions:

  • Clarity of Position
  • Negotiation Strategy: Competing, collaborating, compromising, avoiding, or accommodating
  • Persuasiveness
  • Active Listening

The chatbot is required to deliver detailed feedback with a focus on growth and learning.

6. Ethical Decision-Making Task

This task uses Rest’s (1986) Four-Component Model to assess moral reasoning:

  • Moral Sensitivity: Awareness of stakeholders and consequences
  • Moral Judgment: Balancing competing ethical principles
  • Moral Motivation: Prioritising moral values over gain
  • Moral Character: Integrity and action in ambiguity

The model guides feedback on justifications, confidence, AI influence, and ethical reasoning.

7. Risk-Based Decision-Making Task

Informed by Kahneman and Tversky’s (1979) certainty theory, this task evaluates how participants perceive and respond to uncertainty. The chatbot analyses:

  • Risk preferences and consistency
  • Confidence–risk alignment
  • AI reliance in uncertain contexts
  • Feedback for improving future decisions under uncertainty

Model and Parameters

The underlying chatbot model is OpenAI’s GPT-4o-mini, with a temperature set at 0.6 to balance coherence and creativity in dialogue.

References

  • De Vries, R. E. et al. (2009). Communication Research, 36, 178–206.
  • Finke, R. A., Ward, T. B., & Smith, S. M. (1992). Creative Cognition. MIT Press.
  • Ganesan, S. (1993). Journal of Marketing Research, 30(2), 183–203.
  • Kahneman, D., & Tversky, A. (1979). Econometrica, 47(2), 263–291.
  • Kline, T. J. B. (1996). Personality and Individual Differences, 21(6), 1077–1079.
  • Moreau, C. P., & Dahl, D. W. (2005). Journal of Consumer Research, 32(1), 13–22.
  • Rest, J. R. (1986). Moral Development. Praeger.
  • Sabbah, J., & Li, F. (2025). Innovation, 1–34.