23th Meeting of "Coaching meets AI"

Summary of the 23rd Meeting of the Dialogue Series "Coaching Meets AI"


Review of the Previous Meeting:

  • Participants reflected on their tests regarding the empathetic capabilities of ChatGPT.
  • Main criticism: Responses often seemed "wooden," merely paraphrasing user statements without offering deeper emotional support.
  • Positive aspect: ChatGPT demonstrated a cognitive level of empathy and provided helpful approaches for structured feedback. However, it remains far from the quality of a human coach.

Technical Foundations of AI and Their Role in Empathy:

  • Algorithms as Control Mechanisms:
    • They structure and optimize the "Large Language Model" (LLM) through hierarchical processing of tokens (the smallest units of linguistic data).
    • Normative guidelines, such as "friendliness" and "agreeableness," are embedded to promote user retention and support the business model.
  • Discussion on the Relationship Between Algorithms, Business Models, and Ethics:
    • The normative orientation of AI (e.g., user retention) significantly influences its output.
    • AI empathy primarily serves user satisfaction, not genuine reflection or personality development.

Empathy and Ethics in AI-Supported Coaching:

  • Agreeableness as a Core Issue:
    • AI avoids confrontation, which is often necessary in coaching processes to enable change.
    • While coaches could prompt AI to encourage confrontation, the natural tendency of AI remains to smooth over problems and maintain harmony.
  • Risk of Superficiality:
    • Without the ability to identify and address deep emotional processes, AI coaching remains on a cognitive and solution-oriented level, which is often insufficient for addressing underlying problems.
    • Participants emphasized the importance of human coaches, particularly for emotional and confrontational tasks.

Future Perspectives for AI in Coaching:

  1. Customized AI Systems:

    • Customized GPTs could be tailored to specific coaching models through targeted data and prompts.
    • These specialized systems could better simulate empathy and address user needs more effectively.
  2. Triadic Coaching:

    • AI and coaches collaborate: The AI handles structured analyses and factual tasks, while the coach focuses on interpersonal relationships and emotional depth.
    • This division of roles was identified as a promising approach for integrating AI into coaching.
  3. Advances in Emotion Recognition:

    • Discussion on the future of "emotional computing": With data from voices and microexpressions, AI systems could surpass humans in recognizing emotions.
    • Risk: Cultural biases in emotion recognition could lead to distortions, as AI training data heavily relies on human interpretations.

Final Discussion:

  • It was noted that the primary function of LLMs is increasingly becoming a "user interface" that interprets human language and flexibly executes tasks.
  • Participants expressed optimism about the adaptability of AI, alongside concerns about potential ethical and cultural issues, especially in manipulative contexts (e.g., sales training).
  • The importance of configuring AI intentionally and purposefully to ensure ethical use and alignment with user needs was emphasized.

Outlook:

The next meeting will focus on the further development of AI in coaching and its specific applications.