Bridgelytic Capital ·
From student evidence to product decisions
A project-based AI research method that makes learning, uncertainty and the next product choice visible.
Executive summary
Original Bridgelytic editorial method. Examples are hypothetical, not client results or validated scientific findings.
- Agree a decision, not just a prototype assignment.
- Keep interviews, interpretations and AI suggestions distinct.
- A demo is evidence for discussion, not proof of product readiness.
Give the project a decision boundary
Begin with a question a product owner can act on: which user problem deserves further evaluation? Agree learning objectives, individual contributions, review roles and the limits of the project. Supervised student work can inform a decision without carrying production responsibility.
Listen with permission
Use an interview guide and approved participant process. Explain how notes will be used. Minimize personal information and separate interview observations from the team's interpretation. A small exploratory sample identifies questions, not population-level demand.
Keep competing hypotheses alive
Write at least two plausible explanations of the problem where the evidence permits alternatives. Record supporting observations, contradictions and gaps. AI can organize approved notes, but a human checks every claimed source and decides which interpretation is justified.
Choose an artifact that tests the assumption
A journey map can test task order. A low-fidelity prototype can surface usability questions. A synthetic dataset can test a bounded software behavior. Avoid real personal data when a synthetic input can answer the question. Document accessibility checks and what remains untested.
Review the work, not the confidence of the presentation
Use a progress record, peer critique where agreed and supervisor review. Compare the artifact with the original question. Record failures, disagreements and revisions. Keep a source register alongside the product requirements document so a decision does not outlive its evidence.
Close with a decision memo
Summarize what was observed, inferred and unresolved. Recommend continue, revise or stop, with the next test and its owner. A student demo is not validation of a finished product. Production use requires a separate readiness, permissions and safety review. Portfolio use, attribution and academic authorship follow written permissions.
Evidence to decision
- 01
Question
Agree the learning objective and product decision.
- 02
Listen
Use approved interviews and a source register.
- 03
Compare
Retain alternatives and contradictory evidence.
- 04
Prototype
Test a specific assumption with bounded inputs.
- 05
Decide
Present limits and the next justified step.
Anonymized hypothetical studio illustration
A student team explores an AI-assisted decision experience. Interview notes suggest two different user needs. The team compares journey sketches before building a narrow prototype.
A review may find that the question needs revision rather than more code. This is a hypothetical method illustration, not a named project, institutional partnership or achieved product outcome.
Action checklist
- Document supervision, permissions and contribution boundaries.
- Separate observations, interpretations and generated suggestions.
- Track contradictory evidence and sample limits.
- Test accessibility and error paths appropriate to the prototype.
- Agree portfolio and academic authorship permissions before sharing.
Sources and context
- AI Risk Management Framework
Voluntary risk-management guidance. It supports context, evaluation and accountability, not endorsement or certification of this method.
- Web Content Accessibility Guidelines 2.2
A technical accessibility reference for interface evaluation. A checklist or prototype review does not establish full conformance.