NNaval
← All frameworks
Innovation

Problems, Conjectures, Criticism

Start from a problem, guess a solution, then attack your own guess.

Difficulty
Moderate
Time to result
~ongoing to results
Steps
6
Confidence
86%

Deutsch presents Popper's method as one idea with wide reach: everything starts with a problem, and there is no royal road to solving it. You cannot induce truth by accumulating observations, because every perception is theory laden. Instead you invent a conjecture, an explanatory guess about what is going on, and then subject it to criticism, whether by argument, by test, or by checking whether it even holds together. What survives criticism is provisional knowledge, and it carries new problems that start the loop again. Deutsch adds a footnote to Popper: the point is not testability for its own sake, but explanation. Good explanations are the output, and testability matters only where a field, such as physics, can test explanations against each other.

Origin

Karl Popper's philosophy of science, which Deutsch describes as one idea applied broadly. Deutsch adds his own footnote in this conversation with Naval: that the purpose of the whole exercise is explanation, not merely testability, and that epistemology itself is never a solid foundation but always imperfectly stated and open to improvement.

Core principles

  • 01All progress begins with a problem, never with a blank slate or raw data.
  • 02There is no royal road: no method mechanically generates the right answer.
  • 03Every idea is a creative guess, not something copied from the environment.
  • 04Criticism is the engine, not the brake, of knowledge creation.
  • 05The goal is a better explanation, not a final theory.

How to run it

  1. 1

    Name the problem

    Write down the specific problem you are trying to solve, as a tension between a current explanation and something it fails to account for. Without a stated problem there is nothing for a conjecture to answer to.

    Pro tip Phrase it as a question a child would ask: why does this happen and not that?

  2. 2

    Accept that there is no royal road

    Drop the search for a procedure that grinds out the answer. No amount of data collection, consensus or authority substitutes for inventing an explanation.

    Watch out Waiting for more data before guessing is the most common way to stall indefinitely.

  3. 3

    Make creative guesses

    Propose explanations for the problem. Deutsch stresses these are guesses, not readings taken from nature, and that trial and error, variation and selection, and conjecture and criticism are the same method under different names.

    Pro tip Force out a second and third rival guess so criticism has something to choose between.

  4. 4

    Criticise each guess on its own terms

    Take each conjecture literally and see what it implies. Many collapse under their own consequences before any evidence is needed.

    Pro tip Ask what the theory says about itself, since self-refutation is the cheapest kind of criticism.

    Watch out Rejecting a guess because it is unfashionable or unpopular is social feedback, not criticism.

  5. 5

    Test only where testing discriminates

    Where a field allows it, design a test that separates two surviving rival explanations. If only one explanation is in play, you are demonstrating or measuring, not testing.

    Watch out Genuine tests between rival good explanations are rare, so do not expect every question to be settled by an experiment.

  6. 6

    Treat the survivor as provisional and harvest the new problems

    Keep the best surviving explanation, but expect it to be superseded. Deutsch notes that a superseded theory contains the problems its successor solves, so record what the current answer still fails to explain.

    Pro tip Maintain a running list of unexplained residue; that list is your next research agenda.

    Watch out Declaring a theory final ends the loop and ends the progress.

In the wild

Teaching a six-year-old that everything is a guess

Naval describes teaching the method to his six-year-old son so he has the foundation early. The child now understands intuitively that everything is a guess, so whenever he asks why something happens, the response is not an authoritative answer but an invitation: let us start making some guesses. The conversation then moves to criticising the guesses rather than deferring to what an adult asserts.

The child treats explanations as candidates to be improved rather than facts to be memorised, and asking why becomes the start of a process instead of a request for authority.

Matt Ridley on innovation as trial and error

Naval cites interviewing Matt Ridley, whose books on genomes, optimism and innovation all describe the same mechanism under different labels: trial and error, variation and selection, and in science conjecture and criticism. Once the pattern is absorbed, it shows up everywhere, from biological evolution to technology innovation to art, and in each case the participants are guessing and then filtering, not reading answers off the environment.

One method explains innovation across biology, technology and science, which lets practitioners in one domain borrow the error-correction machinery of another.

Common mistakes

Waiting for the data to speak

Treating observation as the source of theories inverts the method. Deutsch is explicit that perception is theory laden and that we are not copying knowledge from the environment, so a conjecture has to be invented before evidence can bear on anything.

Being a Popperian only in your own field

Deutsch notes that creative people tend to apply the method in their domain and then insist it is a special property of that domain. That parochialism blocks the transfer of error correction to the rest of their thinking.

Chasing a final theory

Aiming at a bucket of finished truths kills the loop, because the surviving explanation is supposed to generate the next problem. Deutsch frames improvement as infinite rather than terminating.

Is it for you?

Best for

Researchers, founders and technologists who need to make real progress on hard, open-ended questions.

Not ideal for

Routine execution work where the problem is already solved and the task is repetition at scale.

From the transcript

popper only had one idea and that is that it all begins with problems and there's no Royal Road to solving them

David Deutsch · (01:00)

if you're going to make progress the only possible way of doing it is finding the problem purported Solutions and then criticizing those Solutions

David Deutsch · (02:00)

these are all just the same method these are creative guesses

Naval Ravikant · (02:30)

From the episode

David Deutsch: Knowledge Creation and The Human Race, Part 2