Effect of AI-Interaction Sequence on Computational Design Outcomes

Determine whether different sequences of AI interactions in AI-enabled computational design science would produce the same design as the AI4CDS process used in the study.

Background

AI4CDS is evaluated through a single computational design project centered on the ChildRiskGuard short-form video safety surveillance artifact. Although the framework organizes AI participation across problem formulation, resource construction, design search, evaluation, and knowledge abstraction, the study does not establish that its particular sequence of AI interactions is necessary or outcome-equivalent to other possible workflows.

The unresolved issue is whether changing the ordering or structure of interactions between researchers and AI systems would lead to the same final computational design. Resolving this question would help distinguish design outcomes attributable to the framework’s substantive principles from those dependent on the particular path taken through the AI-supported research process.

References

The study does not establish that its five phases are the only effective organization of AI-enabled CDS, or that AI4CDS reduces research effort, expands search more effectively than alternative workflows, or produces superior artifacts. It also does not determine whether different sequences of AI interactions would have produced the same design.

— AI for Computational Design Science: A Responsible Human-AI Framework and Case Study on Short-Form Video Safety Surveillance  (2609.05270 - Zhang et al., 4 Sep 2026) in Section 5.5, “Boundary Conditions and Future CDS Research”