Both stepwise constructive-conflict guidance and the antagonistic agent led novice designers to reconsider what makes a design effective more than unsupported self-reflection¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r2 · q3 · i3 large
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i3 On the reconsideration item, both the Stepwise Guidance and Interactive Engagement groups scored significantly higher than Self Reflection, with a large omnibus effect (ε² = .19). → Howard Ziyu Han 2026 (2)

Evidence¶

Howard Ziyu Han 2026¶

Howard Ziyu Han, Nikolas Martelaro. (2026). Enacting Constructive Conflicts with AI Agents to Enhance Reconsideration among Novice Interaction Designers. https://arxiv.org/abs/2608.04166 doi:10.48550/arxiv.2608.04166

q3 · i3 · causal · r2

Between-subjects experiment with 45 design students in three conditions; Kruskal–Wallis on the post-survey item showed "significant group differences" with the printed omnibus effect size ε² = .19, a large effect.

"On the critical thinking item“To what extent do you reconsider what makes a design effective in this challenge?”, significant group differences were found (𝐻( 2)= 8.30, 𝑝=. 016,𝜀2 =. 19, large effect)."

Howard Ziyu Han 2026 (2)¶

Howard Ziyu Han, Nikolas Martelaro. (2026). Enacting Constructive Conflicts with AI Agents to Enhance Reconsideration among Novice Interaction Designers. https://arxiv.org/abs/2608.04166

q3 · i3 · causal · r2

Post-hoc pairwise contrasts from the same experiment: both guided groups scored higher than Self Reflection with Hedges' g = 0.93, a large effect for each contrast.

"Dunn’s post-hoc tests with Bonferroni correction showed that both the Stepwise Guidance group (𝑀= 5.47,𝑆𝐷= 1.246;𝑝𝑎𝑑 𝑗 =. 048, Hedges’ 𝑔= 0.93, large ef- fect) and Interactive Engagement group (𝑀= 5.47,𝑆𝐷= 1.252; 𝑝𝑎𝑑 𝑗 =. 030,𝑔= 0.93, large effect) scored higher than the Self Re- flection (𝑀= 4.00,𝑆𝐷= 1.77)."

Discussion¶

Related Claims¶

• Reflective Practice Improves Outcomes When Structured — the general claim this finding bears on, with the studies behind it

• Conflict-based instruction improves science conceptual learning, though staged contradictions helped only learners who reported being confused, and no study isolates disequilibrium as the mechanism — the general claim this finding bears on, with the studies behind it

• Only the interactive antagonistic agent, not stepwise guidance, significantly shifted novice designers' design thinking relative to self-reflection — related
• Both constructive-conflict conditions produced substantially more idea editing between design iterations than unsupported self-reflection — related
• Novice designers deleted ideas only when confronted with stakeholder friction; unsupported self-reflection produced no deletions — related
• Stepwise constructive-conflict guidance reduced new idea generation, while the antagonistic agent played a more balanced role — related
• Participants valued the AI as a relational buffer that softened interpersonal pressure when disagreement required concession, without expecting it to resolve conflict — related
• The antagonistic agent surfaced a broader set of stakeholder concerns than novice designers reached on their own — related