DATATOMIC / METHOD
Start with observation.
Follow the question.
An empirical starting point, informed by grounded theory. Observations shape the explanation; mathematics makes its assumptions and consequences explicit.
MORE THAN ONE WAY TO REASON
Induction helps identify patterns. Abduction develops plausible explanations. Deduction asks what should follow if an explanation is correct.
We move between these modes as the question demands, returning to observations whenever a model needs to be challenged.
- 01
Observe
Begin with a phenomenon, a discrepancy, or a pattern that deserves a closer look.
- 02
Find the pattern
Compare observations. Look for relationships, exceptions, and competing interpretations.
- 03
Propose explanations
Develop more than one plausible explanation and identify what each assumes.
- 04
Formalize
Use mathematics or a computational model to make the idea precise and derive its consequences.
- 05
Test
Use measurements, simulations, or prototypes. Compare with a baseline and look for evidence against the proposal.
- 06
Revise & share
Refine or abandon the model. Document assumptions, results, limitations, and the next question.
WHEN THE MACHINE CANNOT YET BE BUILT
Make the assumptions visible.
Mathematics can expose contradictions, establish limits, and specify what a future experiment would need to measure. Simulations examine consequences under chosen assumptions. Evidence from the physical world is needed to establish whether those assumptions describe reality.
RESEARCH STAGES
Say where the work stands.
Question
A phenomenon and a bounded question worth investigating.
Model
A proposed explanation with assumptions and derived consequences.
Simulation
A computational experiment under specified conditions.
Prototype
An implementation that makes the proposed mechanism observable.
Empirical result
Measurements with a baseline, uncertainty, and stated limitations.
These stages describe different kinds of work. A project may revisit an earlier stage as evidence changes.
THE PRACTICAL TOOLKIT
Tools follow the question.
01 / DISCIPLINE
Artificial intelligence
Systems that retrieve, reason, and work with language.
- LLM systems
- Agentic workflows
- Retrieval systems
- Machine learning
- Natural language processing
- AI-assisted applications
02 / DISCIPLINE
Data + intelligence
Find structure in information. Keep the evidence in view.
- Data engineering
- Analytics
- Signal extraction
- Alternative data
- Visualization
- Information fusion
03 / DISCIPLINE
Quantitative systems
Make assumptions explicit and put models to the test.
- Financial modeling
- Algorithmic analysis
- Market intelligence
- Backtesting
- Risk analysis
- Signal development
04 / DISCIPLINE
Software engineering
Turn research into systems people can actually use.
- Web applications
- APIs
- Automation
- Data pipelines
- Rapid prototyping
- Product engineering
05 / DISCIPLINE
Cyber + intelligence
Understand digital behavior through structured investigation.
- Cybersecurity research
- Threat intelligence
- Malware and data analysis
- Structured intelligence workflows
06 / DISCIPLINE
Experimental R&D
Build just enough to find out what is possible.
- Prototype development
- Technical feasibility
- Emerging technologies
- New interfaces
- New product concepts
METHOD IN CONTEXT
The grounded-theory influence comes from its iterative relationship between observations and developing theory. Physical and computational questions also require methods appropriate to their evidence.
Grounded theory · Kathy Charmaz & Antony Bryant