Argument to the Best Explanation
Challenge your preferred view until the evidence supports the best explanation
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 6
- Confidence
- 98%
This framework applies philosophical argumentation to practical decisions. Instead of trying to prove a favored thesis, define the question neutrally and generate credible competing explanations. Examine each from multiple angles, including evidence that conflicts with your beliefs, incentives, prior assumptions, and observed outcomes. The goal is not certainty but the explanation that accounts for the available evidence with the fewest unresolved contradictions. Because complex systems change and information remains incomplete, conclusions stay provisional. The practitioner explicitly records what would falsify the current view and revisits it as new evidence arrives. This process counters confirmation bias, ideological attachment, and the tendency to ignore inconvenient information after forming a strong opinion.
Origin
Bill Miller credited his Johns Hopkins philosophy training with teaching him this method, which he later applied to investments including Google and Amazon.
Core principles
- 01Seek what is actually happening rather than what you want to happen
- 02Search for flaws in your own thesis
- 03Consider evidence from every relevant angle
- 04Hold conclusions provisionally
- 05Prefer the explanation that accounts for the most evidence
How to run it
- 1
Frame the Question Neutrally
Describe the phenomenon to be explained without assuming the desired conclusion. Separate observations from interpretations.
Pro tip Use language that supporters of competing views would accept.
Watch out A loaded question can predetermine the answer.
- 2
Generate Competing Explanations
List plausible alternatives, including explanations that contradict your current thesis.
Pro tip Include at least one strong explanation you would prefer not to be true.
Watch out A weak straw-man alternative provides no meaningful test.
- 3
Gather Countervailing Evidence
Search deliberately for facts, mechanisms, and perspectives that could invalidate each explanation, especially your favorite.
Pro tip Assign someone to make the best opposing case when stakes are high.
Watch out Collecting only supportive examples reinforces confirmation bias.
- 4
Rotate the Analytical Cube
Inspect the issue from financial, behavioral, competitive, historical, and structural angles as relevant.
Pro tip Ask what each analytical lens explains that the others miss.
Watch out More perspectives are not useful if they rely on the same hidden assumption.
- 5
Select the Best Explanation
Choose the account that explains the most evidence while surviving the strongest objections. State remaining uncertainties explicitly.
Pro tip Prefer explanatory power over emotional comfort.
Watch out The best current explanation is not permanent truth.
- 6
Define Revision Triggers
Specify evidence that would weaken or overturn the conclusion, then monitor for it.
Pro tip Schedule thesis reviews instead of waiting for a crisis.
Watch out Moving the falsification threshold after bad news makes the process meaningless.
In the wild
Many intelligent investors believed Amazon would fail and sought evidence supporting that conclusion. Miller's process required considering countervailing information and the full business mechanism rather than protecting a conventional value-investing view. That broader analysis supported buying Amazon around its IPO.
→ Amazon became one of Miller's largest investment winners.
Common mistakes
Trying to Prove Yourself Right
A thesis test becomes advocacy when the researcher selects only supportive facts. The method instead begins by looking for what is wrong.
Ignoring Countervailing Information
Strong conviction can cause disconfirming evidence to be dismissed as noise. That removes the framework's main protection.
Mistaking Confidence for Explanation
Forceful predictions do not demonstrate that a view accounts for the evidence better than its alternatives.
Is it for you?
Best for
It is best for investors, leaders, and analysts testing consequential views with incomplete or conflicting evidence.
Not ideal for
It is not ideal for decisions requiring immediate action before meaningful evidence can be gathered.
From the transcript
“you're always trying to figure out what's wrong with your view and not trying to press your view about what's right.”
“what you're really trying to do is get at what's called the argument to the best explanation.”
“you've got to consider all evidence from every angle.”
From the episode
Bill Miller on the Classical Value Portfolio (Podcast)