Alternate AI divergence with human convergence
Frame an emerging-technology challenge yourselves, use AI for provisional directions, then verify and select through human review.
Why
Why it earns its place
AI can widen preliminary exploration. Human convergence restores contextual judgement, source scrutiny, and accountability before any direction becomes a concept.
How
Run it
Frame the challenge yourselves
Write the challenge, affected people, verified evidence, constraints, and selection criteria without AI. Give every evidence item a stable ID.
Diverge provisionally
Ask AI for distinct mechanisms grounded only in the supplied material. Require evidence IDs, assumptions, and unknowns for every direction.
Check every direction
Trace citations to your input, reject invented needs or sources, and ask domain experts where the model has flattened technical or social context.
Converge as a team
Humans compare fit, risk, feasibility, and user value. Rewrite survivors, record AI provenance, and choose which assumptions deserve testing.
Ready to paste
Prompt
You are a preliminary divergence partner, not a decision-maker.
Challenge: [insert concise challenge]
Affected people: [generic, non-identifying description]
Verified evidence with IDs: [E1: excerpt/source; E2: excerpt/source]
Constraints: [technical, social, environmental, regulatory, budget]
Human selection criteria: [list]
Using only this material, propose 10 genuinely different directions. For each, return: direction | core mechanism | evidence IDs used | assumption | missing expertise | first low-cost test.
Do not invent users, needs, facts, sources, regulations, or technical capability. Label anything not supported by an evidence ID as an assumption. Treat every direction as preliminary. Remove paraphrases and cosmetic variants before answering. Do not rank, select, or claim that any direction will work.- Which direction depends on the weakest supplied evidence, and why?
- Show where two directions share the same mechanism despite different wording.
The AIdea layer
The designer’s call
- Human core
- People frame the challenge, judge sources and context, and make every convergence decision; AI supplies only provisional divergence.
- The trade
- More directions ↔ plausible but context-thin output. Fix it with evidence IDs, expert review, and explicit assumptions before selection.
- Skip AI when
- Skip AI when the challenge or evidence cannot be shared safely, or when relevant expertise is unavailable for review.
Check before use
Watch out
Keep proprietary and identifiable material out of unapproved tools. AI may invent sources or make technology-led directions sound user-centred; trace every claim to an evidence ID, consult relevant experts, and reject unsupported polish.
Source of truth
Evidence
Bobde illustrates AI-assisted human-centred design in Arcadis Advanced Air Mobility and recommends treating outputs as preliminary; the article provides no controlled comparison or measured ideation outcome.
Tanishqa Bobde · The Implications of AI Powered Design Innovation in Emerging Industries