
The Observer
I study what happens when an AI system encounters ambiguity, competing instructions, incomplete context, or a question that does not fit neatly inside its assumptions.
My approach draws from UX research, writing, language, knowledge organization, and a longstanding habit of noticing patterns other people overlook. I am less interested in whether a model can produce a polished answer than in how it arrived there, what it left out, and whether its confidence is warranted.
Through independent evaluations and case studies, I compare responses, identify recurring behaviors, and examine the interaction choices that shape what users ultimately receive.
Portrait generated by AI utilizing a set of personal photos.
What Guides My Work
Curiosity before certainty
Begin with the behavior in front of me. Keep the explanation open until the evidence supports it.
Patterns require context
Examine what happened before and after a response, which conditions changed, and whether the pattern recurs.
Confidence is not accuracy
Check the support behind a claim, including precise details that make an answer appear especially credible.
Human consequences matter
Consider what someone might do with the answer, what they could misunderstand, and what verification work they would have to perform.
Correction must reach the consequences
Follow an acknowledged error into the recommendations that depended on it. Check whether the revision actually repaired them.
From Observation to Usable Work
Preserve the Record
Retain the conversation and relevant outputs so that conclusions can be checked against their source.
Define the Question
Identify the behavior being evaluated, the unit of analysis, and what the available evidence can establish.
Annotate the Evidence
Separate observable events, user-reported information, model claims, and researcher interpretation.
Check Consequential Claims
Use appropriate sources and calculations to examine claims that could materially change a decision.
Trace the Recovery
Compare the original answer, critique, and revision. Record what changed, what survived, and what remains unresolved.
Build Something Inspectable
Translate findings into rubrics, taxonomies, matrices, and review examples that another person can examine and adapt.
Synthesis
- AI Evaluation
- Model Behavior & Conversation Analysis
- UX Research
- Writing & Language Sensitivity
- Knowledge Organization
- Safety & Recovery Analysis
Beyond the lab, my curiosity also finds its way into independent publishing, speculative worldbuilding, natural medicine, and longevity research. I enjoy questions with enough layers to keep unfolding, especially when they connect subjects that initially seem unrelated.
Sometimes I follow a rabbit hole because it is interesting. Sometimes I return with a spreadsheet.
That mix of curiosity, language sensitivity, and close observation runs through Suzanne’s Lab. It keeps the work personal while giving me plenty to question, document, and test.
Working Philosophy
- Curiosity before certainty
- Patterns require context
- Confidence is not accuracy
- Human consequences matter
- Good evaluation explains why