Problem-First Epistemology
Start every inquiry from a clash of ideas, not from a search for certainty
- Difficulty
- Moderate
- Time to result
- ~ongoing to results
- Steps
- 6
- Confidence
- 78%
Deutsch reframes the growth of knowledge as problem-driven rather than justification-driven. The input is two or more ideas making incompatible claims about the same thing, which is what a problem actually is: a clash. The process is to accept that they cannot both be true, to accept that both may well be false, and then to criticise, vary and correct the errors in each. The output is a better explanation plus a new and better problem. The key move is dropping the question 'how do I know this is justified?', which forces a regress back to gods, senses or pure reason, and replacing it with 'what problem does this solve, and where does it go wrong?'. Because clashing ideas quietly change each other, even an argument nobody wins leaves both parties holding better ideas than they walked in with.
Origin
Deutsch credits Karl Popper with the concept of a problem as the seed of knowledge, and says a colleague convinced him it is Popper's most important contribution, more important than falsifiability. Extracted from Naval.
Core principles
- 01The growth of knowledge begins with problems, not with observations.
- 02A problem is a clash between ideas that cannot both be true.
- 03Both clashing ideas are usually false; the prize is error-correction, not victory.
- 04Asking where knowledge came from starts an infinite regress; asking what it explains does not.
- 05Even an unresolved argument changes both ideas for the better.
How to run it
- 1
Surface the clash
Write down the two ideas, interpretations or theories that are making competing claims about the same thing. If you cannot name two, you do not yet have a problem, you have a topic.
Pro tip The clash can be between two theories, or between a theory and a theory about an observation. Observations are never raw; they carry theory inside them.
- 2
Refuse the truce
State plainly that both cannot be true. The refusal to blur them is what turns a vague discomfort into a workable problem.
Watch out Reconciling two incompatible claims with a diplomatic 'both have a point' destroys the problem and stops the inquiry.
- 3
Assume both are false
Treat the default case as both sides being wrong, with the more important errors concentrated in one of them. This stops the search collapsing into picking a side.
Pro tip Ask which errors would matter most if they were corrected, and start there.
- 4
Criticise for explanatory quality, not for pedigree
Judge each idea by how hard it is to vary while still explaining what it explains. Do not judge it by its source, its author's status or how certain its proponents sound.
Watch out The moment an argument turns on someone's credentials or honesty, the epistemology has already failed.
- 5
Let the argument run even if nobody concedes
Keep the ideas in contact. Both sides will find new angles they did not hold before the exchange, which changes the ideas whether or not anyone changes their mind.
Pro tip Capture the new angle you found while defending your own view; that is the actual output of the argument.
- 6
Bank the successor problem
Record the new problem your improved explanation has opened up. A solved problem that generates no successor usually means the explanation was shallow.
In the wild
The anomalous orbit of Mercury is usually told as an observation that refuted Newton. Deutsch retells it as a field of clashing ideas: Newton's theory, Einstein's theory, theories about whether the astronomers' observations were even correct, and rescue theories positing an unknown planet at a specific mass and position. Each rescue could be pinned down and ruled out. Critics said Eddington's eclipse observations were wrong, and only much later were they uncontroversially confirmed. The argument improved every theory in play until only one good explanation was left standing.
→ General relativity survived not because one observation killed Newton, but because criticism eliminated every competing good explanation and left an infinity of merely bad ones.
Deutsch uses religious justification as the cautionary case: the claim that Jesus is the son of David is defended by going back to first principles, which requires a genealogy stretching back further than anyone has ever documented, so one gets invented. The more important the belief, the harder people work on its pedigree instead of its content. Deutsch calls this a simple epistemological error that has led to unlimited suffering, with people torturing and warring over interpretations of what someone may or may not have said thousands of years ago.
→ Effort flows into manufacturing a foundation rather than correcting errors, and the belief becomes untouchable rather than better.
Common mistakes
Treating every problem as a setback
In this sense a problem is a good thing: an interesting gap that has not been closed yet. Reading the word as a threat makes people hide clashes instead of hunting for them.
Trying to find the true one
Framing the clash as 'which of these is correct?' assumes one of them is. Usually both are false and the real work is locating the errors that matter most.
Auditing the source instead of the errors
Demanding to know where an idea came from before engaging with it triggers an endless regress and shifts the argument onto authority. Judge the explanation, not the pedigree.
Is it for you?
Best for
Researchers, analysts and founders who need a way to open and run a genuine disagreement so that it produces better explanations.
Not ideal for
Situations that need a fast decision under a deadline rather than a deeper understanding of the underlying conflict.
From the transcript
“if I had to pick something that most people don't get about epistemology it's that the growth of knowledge begins with problems”
“thinking of a problem as a clash and it's got to be a clash of ideas or interpretations or theories or so on is Illuminating”
“a clash of ideas is very beneficial even if they are never resolved”
From the episode
The Deutsch Files II