Approach & tools
Before the answer, examine the question.
“Every problem is a technical problem” begins, in my work, with examining how the question is defined. Here is how I ask questions, test my own explanations, and choose tools to make an idea inspectable.
Three things help us start.
What you want to achieve
Who will use the result? What decision comes next? What would make this useful?
What you have already tried
Existing plans, observations, data, or a prototype help me understand the problem.
What we can work with
Time, budget, access to users, available tools, and the people who need to take part.
Reframing changes what we can do next.
These are shifts I draw from my practice. A new question must still serve the original goal. Being easier to answer is not enough.
- Frontstage · ExperienceStart, perform the task, end
- Backstage · SimulationA person follows response rules
- Observe & decideTest understanding before implementation
Observe understanding · Simulate the responseConceptual model
Robot interaction
Starting questionHow do we test an unfinished feature?
Separate understanding from implementation.
Simulate key responses with a researcher. Observe how people begin, complete, and end the interaction.
Read the case- Source recordsKeep tasks and context
- Tasks & observationsSeparate actions and interpretations
- Traceable issuesEach finding has a source
Records → Tasks & observations → Traceable issuesConceptual model
Research records
Starting questionHow do we organize all these records?
Ask which decision needs which evidence.
Locate observations by task, connect each issue to its source, then discuss what to change.
Read the case- InsightDefine the assumption to examine
- Design ⇄ ObserveReturn feedback while changes are possible
- ConfirmationCheck whether the question was answered
Insight → Design & observation in a loop → ConfirmationConceptual model
Design validation
Starting questionHow do we test once development is done?
Check while changes are still possible.
Observe use during design and implementation. Feed findings into the next version, then plan confirmation.
Read the caseMetacognition: examine how the question is framed.
01
Separate goal from proposal
“Hire more people,” “run training,” and “add AI” are proposals. What change do we actually want, and how will we know it happened?
Output: an agreed question, scope, and criteria for judging the result.
02
Inspect conditions and relationships
What happens across people, information, tools, and handoffs? Is the obstacle within a step or between steps? Which constraints are fixed, and which are habits?
Output: observations, possible causes, and the evidence we most need next.
03
Test competing explanations
Did the person misunderstand, or did the system fail to respond? Use a simulation, a recording, or a script to look for an observable difference.
Output: a study plan, simulated experience, or tool prototype we can inspect.
04
Check my own reasoning, too
What result would change my mind? Does the reframed question still serve the original goal? What did the intervention change? Then continue, adjust, or stop.
Output: results, instructions, open issues, and a plan for further checks.
Toolbox / Methods I have used
The problem determines the tool.
My toolbox includes measurement equipment, research methods, code, and AI. These are concrete jobs they have done. A new problem calls for a fresh choice.
- OBSERVATIONEye trackingEEGBehavior
Signals in the context of use
See what actually happens
Eye tracking · EEG · Interviews · Observation
In packaging research, I examined visual paths, compared early responses, and observed handling and opening. I interpreted each within the same stage of use.
Packaging research - PROTOTYPINGWizard-of-OzTask scriptsSimulation
Test assumptions through experience
Try an interaction before building it
Wizard-of-Oz · Task scripts · Simulations
In service-robot research, a person behind the scenes simulated unfinished responses so we could observe how users started, completed, and ended an interaction.
Service-robot research - RETRIEVALPythonChromaDBSemantic search
From a question to sourced context
Find context that answers a question
Python · ChromaDB · Keyword and semantic search
I maintain separate product and usability libraries, adjusting scope and passage boundaries so retrieval returns relevant context with a route back to the source.
Personal knowledge and AI system - ORCHESTRATIONOpenClawClaude CodeInstructions
Bounded execution, checkable results
Connect repeated steps
OpenClaw · Claude Code · Reusable task instructions
I combine a runtime, coding assistant, and existing tools for lookup, information handling, and filing, defining inputs, permissions, and checks for everyday use.
Daily AI workflows
Before choosing a tool, I ask which uncertainty it can resolve or which obstacle it can change. If an observation or a table is enough, I start there.
Check whether the approach helped before investing further.
Research needs its source records, design recommendations need a follow-up check, and tools need real tasks. If evidence is missing, I explain the gap. A small check may be enough to answer the question; further investment follows from what remains worth solving.
Read the casesMaterials to use in your own work
These blank templates and instruction examples were prepared for this website. Download and adapt them to your task. They contain no client data and are not original project deliverables or evidence of results.
- Blank template
Observation and issue traceability
Record actions, interpretations, sources, and pending decisions separately.
- Blank template
Study preparation and decisions
Define assumptions, tasks, recording methods, and who acts on findings.
- Blank template
Agent human-workload evaluation
Compare setup, supervision, correction, and recovery; includes a blank CSV.
- Instruction example · Awaiting task validation
Evidence-linked research brief Skill
A minimal instruction example with inputs, evidence records, exceptions, and human review.