TRAIF2026—Delvina L. Morrow

Value of the Information & Strength of the Argument

The talk offers a compelling synthesis of existing research and a clear conceptual contribution: distinguishing ‘responsible AI’ (preventing harm) from ’empathy by design’ (actively cultivating values). The argument is logically structured, moving from empirical evidence (SEL effectiveness, empathy gap) to practical design insights (value alignment from inception, stakeholder co-design). The use of personal narrative (spaghetti story, daughter Parker) humanises the argument and effectively conveys urgency. However, the argumentation relies heavily on anecdotal evidence and the author’s own claims about GLEA’s benefits; no rigorous outcome studies are presented. The leap from a single classroom anecdote to ‘kind cities’ remains aspirational, but the reasoning is clear and the evidence used to support the framework is credible.

Scientific Rigor, Source Quality, Title Accuracy

Scientific rigour is moderate: the talk cites several high-quality sources (CASEL, Durlak et al. meta-analysis, Cambridge’s Nomisha Kuran, Torres in Scientific Reports) and correctly summarises their findings. The description links provide institutional references (IEAI, alignAI, Responsible AI Forum) but these are primarily about the conference, not the specific claims. No URLs are given for the cited papers, but they are identifiable from the transcript. The title is the event and speaker name, offering no content information; this is a missed opportunity for discoverability. The adequacy between title and content is low but not misleading. There are no obvious distortions of cited research; the speaker is careful to note evidence limits (e.g., ‘anecdote, not an outcome study’).

The title simply names the event and speaker, giving no indication of the content. This is common for conference recordings, but fails to convey the topic of empathy-by-design in AI for children.

7/10

The talk is well-structured, references multiple credible studies (CASEL, Durlak et al., Torres 2026, Cambridge empathy gap) and provides a coherent conceptual framework. However, its reliance on anecdotal evidence (e.g., classroom example) and lack of empirical outcome data from GLEA itself limit the scientific rigour. The author's dual role as founder of GLEA introduces potential bias, though transparency is maintained.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:30Spaghetti story: introduces the challenge of defining 'ready' for children and AI.
• 02:20Definition of empathy by design vs. responsible AI; value alignment as proactive vs reactive.
• 03:50Introduction to SEL and CASEL's five core competencies; empathy in social awareness.
• 05:20Empathy gap: the Alexa incident and the need for design that accounts for emotional nuance.
• 07:10Evidence from SEL meta-analyses (Durlak et al.) and the transfer problem: children struggle to apply skills.
• 09:30Introduction of GLEA: mission, target age, use of LLMs, and the KIND framework.
• 11:40Design insights: value alignment from inception and stakeholder co-design.
• 14:10Kindness infrastructure tiers (families to cities) and LEED analogy - 'LEED for kindness'.
• 16:40Call to action: inviting collaboration and stressing deliberate value choices.

• alignAI EU project — Mentioned in description as funding source for the Responsible AI Forum 2026.
• IEAI - Institute for Ethics in Artificial Intelligence — Hosting institution for the forum and the speaker's affiliation context.
• IEAI newsletter — Link provided for further information and updates.
• Responsible AI Forum — The conference where the talk was presented.
• Amerikahaus — Venue and video footage provider.
• IEAI events — Information about other IEAI events.

• Durlak, J. A. et al. (2011). The Impact of Enhancing Students' Social and Emotional Learning: A Meta‐Analysis of School‐Based Universal Interventions. Child Development. — Cited in the talk as evidence for SEL effectiveness across 213 studies.
• Torres, N. (2026). Integrating social-emotional learning into AI architectures improves performance. Scientific Reports. — Cited as showing 49-66% improvement when SEL is integrated as a core design principle.

• Kuran, N. (2023). The empathy gap in AI: how it misleads children. University of Cambridge.