Reflecting Human Values in XAI: Emotional and Reflective Benefits in Creativity Support Tools

Current AI evaluation methods miss important human benefits: The paper's review of creativity support tools found that most evaluations focus on task performance rather than how tools affect users' emotional well-being, self-reflection, or personal growth—suggesting we're measuring the wrong outcomes when assessing whether AI systems are truly beneficial.

Explainable AI should be evaluated for emotional and psychological impacts: Beyond asking "does the AI work?", researchers should measure whether XAI systems help users feel more confident, engaged, or self-aware when using them, particularly in creative domains where these human factors matter significantly.

Four user-centric measures deserve attention in AI design: The paper identifies specific dimensions worth evaluating—developing intrinsic abilities (building user skills), emotional well-being, self-reflection, and self-perception—suggesting that responsible AI should support human flourishing, not just task completion.

The arts reveal gaps in how we evaluate AI: By examining creativity support tools specifically, the paper highlights that general AI evaluation frameworks may be inadequate for domains where human expression, growth, and satisfaction are central goals rather than peripheral concerns.

Responsible AI requires rethinking success metrics: The paper challenges the field to move beyond traditional performance measures and consider whether AI systems align with human values like autonomy, emotional health, and personal development.