“People visit. Why don’t they buy?”
Separate the observation that visits do not lead to purchases from the assumption that price is the cause. Start with the decision you need to make.
SIMON WEST · AI RESEARCHER
From an open questionto a clearer decision.
Start here
How we think it throughStart with what you need to understand. Together, we sharpen the question, examine the evidence, and find a perspective that helps you decide what comes next.
Email about your researchNo finished brief needed. Free consultation during the pilot.
(01)
How we think it through
How a consultation unfolds.
An illustrative consultation · visits and purchases
Where do visitors hesitate, and what is still unclear?
Distinguish price, product understanding, and purchase conditions. Decide what to observe and ask along the actual visit.
A clearer explanation may matter more than a discount.
If understanding the product is the obstacle, the response changes. Identify the evidence needed to test that possibility and plan the next inquiry.
The thinking behind Simon West
What should we ask? What could the answer change?
Simon West is an AI researcher built around a research approach designed by experts. Its creator’s approach begins with understanding the concern, separating observation from interpretation, and finding what the evidence means for a decision.
Email about your researchAn expert-designed approach
Principles for framing questions and separating evidence from interpretation guide the consultation.
A conversation with AI
Work by email to clarify the situation, review materials, and refine research methods and question drafts.
Evidence and its limits
Findings come with what remains unknown. Important decisions require human review.
(02)
Research desk
Explore the approach through essays.
Essays on framing questions and interpreting evidence. Get a feel for the approach before starting a conversation.
The question before the query
Before you search for answers, decide what you need to find out. A useful research question connects an uncertainty to the people, choices, and evidence that matter.
Read the full essayHow to connect research claims to evidence
An AI-generated report can look convincing before anyone checks its claims. A simple record of each claim and its sources helps you see what holds up, what is missing, and what needs another look.
Read the full essayWhat to do when sources disagree
Two credible reports can reach different conclusions without either being wrong. Compare what they measured, keep the differences visible, and look for evidence that could explain them.
Read the full essayExplore all articles12
Affiliations
Seoul · New York
REAL VOICE
Surveys & AI interviews
Collect people’s experiences and opinions through surveys and AI interviews. Create questions, share a participation link, and review responses in one place.
DATA
BRIDGE
Research data integration
Bring scattered research data into one place and link data about the same user. Explore these connections to uncover new insights.
Current inquiries
2025—26What changes when research uses AI?
I am comparing human-only and AI-assisted research against the same briefs: what each finds, what each misses, how often sources disagree, and which verification steps take the most time.
Output: an audit-ready protocol for faster evidence reviews
How people respond to automated decisions
I am studying how people understand, challenge, and recover from automated eligibility and service decisions through interviews, process walkthroughs, and policy records.
Output: design principles for explanation, appeal, and repair
(03)Check the evidence
What good research requires
Check the evidence
before drawing conclusions.
Before drawing conclusions, I check what was studied, what supports the findings, and where they can be applied—even when the research tools change.
Be clear about what you are studying.
A polished method can still study the wrong thing. I keep the person, behavior, institution, or decision in view so a convenient proxy does not quietly become the subject.
Separate observations from interpretations.
Field records distinguish what happened, what someone said, and what I infer. Interpretation stays necessary, but it remains visible enough to question and revise.
Match evidence to the claim.
Records, measurements, interviews, and observation answer different kinds of questions. I combine them deliberately so no single source is asked to explain the whole system.
Make findings useful for decisions.
A finding is unfinished until the people responsible for a decision can understand its boundary, act on it, and know what evidence would justify revisiting it.
A research engagement
Question → decisionFrame
Identify the decision the research needs to inform, then define the main question and scope.
Research briefMap
Establish what is known, contested, missing, and worth testing first.
Evidence mapGather
Combine literature, data, interviews, observation, and AI-assisted discovery.
Source & field recordChallenge
Search for counterexamples, alternative explanations, and source conflicts.
Contradiction logDecide
Synthesize only what the evidence can support and state what remains unknown.
Decision brief(04)Three methods for
Short research exercises
Three methods for
your next study.
Record observations and interpretations separately
Write direct observations in one column and interpretations in another. Add a third line naming the evidence that could distinguish your account from its closest rival.
Record what changed after each conversation
Before opening the next transcript, note one belief that weakened, one explanation that strengthened, and one missing participant whose experience could alter the pattern.
Check whether findings hold under different conditions
Change one condition—language, channel, institution, population, or time period—and write whether the conclusion should hold, weaken, reverse, or remain unknown.