Good explanations cite difference-making causes at the right level of abstraction
Strevens on Explanatory Depth
The Core Idea
Michael Strevens argues that good scientific explanations donât just cite causesâthey cite the difference-making causes while abstracting away irrelevant details.
The Kairetic Account
Strevensâ âkaireticâ account (from the Greek word for âdecisive momentâ) says:
An explanation should include:
- All and only the difference-makers
- The causal factors that make a difference to the outcome
- Abstracted to the right level of generality
Why This Matters
Against Simple Causation
Not all causes belong in explanations. The presence of oxygen is causally necessary for a fire, but we donât cite it when explaining why a particular fire startedâunless oxygen levels were the difference-maker.
Against Pure Laws
Laws alone donât explain. âObjects fall at 9.8 m/s²â doesnât explain why this particular object fellâwe need initial conditions and to show why they mattered.
The Right Level of Abstraction
Good explanations abstract away details that donât make a difference. We explain genetic inheritance without citing quantum mechanics, not because QM is wrong, but because itâs not difference-making at that level.
Application to Research
This framework helps evaluate explanations:
- Am I citing all the difference-makers?
- Am I including irrelevant causal factors?
- Am I at the right level of abstraction for my explanatory target?
Tensions and Questions
- How do we identify difference-makers before we have a full causal model?
- Does this work for statistical explanations?
- What about historical explanations where everything matters?
Connection to My Work
Strevensâ framework shapes how I think about:
- Building computational models (what to include?)
- Writing up results (what level of mechanism to cite?)
- Evaluating theories (are they at the right grain?)
Key Sources
- Strevens, M. (2008). Depth: An Account of Scientific Explanation
- Strevens, M. (2004). âThe Causal and Unification Approaches to Explanation UnifiedâCausallyâ