Naive Rationalism, Anti-Rationalism, and Mature Rationalism

Note

Ozzie Gooen · November 2020 · Completeness: ★☆☆ (sketch)

Unpublished QURI draft, posted as-is — rough and unfinished. Nuño Sempere’s comments are restored at the end (with permission); other collaborators’ comments have been removed.

Ozzie Gooen’s take

It seemed interesting to imagine a clear philosophical tradition emerging around the Berkeley rationality scene and superforecasting. There’s a long tradition of backlash against technocratic and Enlightenment thought; I think there are ways to fix these kinds of thinking rather than throw them out.

Opus 4.8’s summary

A November 2020 draft, essentially a single comparison table preceded by three short definitions. It distinguishes “Naive Rationalism” (Enlightenment/high-modernist overconfidence in quantification and reason), “Anti-Rationalism” (Romantic, Continental, and postmodern rejections of that project), and “Mature Rationalism” (described as current EA/rationalist best practice). The table contrasts the three positions across rows such as expected-value calculations, p-values, RCTs, quantitative and qualitative measurement, ontology, central planning, morality, and rational-agent models. Several cells, particularly under RCTs and qualitative evaluation, are left blank, and the piece offers no surrounding argument beyond the definitions. Restored comments from Nuño Sempere correct an attribution (Platonic versus Aristotelian) and discuss placing Wittgenstein among the anti-rationalists.

Naive Rationalism

By “Naive Rationalism”, I am referring to a long line of thinking that was particularly notable in the Enlightenment but has kept with us. It sometimes goes by the names of “Modern Thought”, “High Modernity”, or “Rationalism.”

Anti-Rationalism

By this, I mean schools of thought that primarily counter those of naive rationalism. These kinds of thinking were popular in Romantic Philosophy, Continental Philosophy, and Post-Modern Philosophy. The Counter-Enlightenment is perhaps the most obvious example. https://en.wikipedia.org/wiki/Counter-Enlightenment

Mature Rationalism

By this, I mean our current best understandings of how to successfully continue the “rationalist project.” This is the current best practices in the EA/rationalist communities.

	Naive Rationalism / Modern Thought	Anti-rationalism	Mature Rationalism
Expected Value Calculations	Independent expected value estimates should be used to decide all things.	EV estimates are totally missing the point. You should go with your gut.	Most EV estimates are highly sensitive to sensitive parameters. They need to be used with caution and ideally created by multiple people independently with lots of verification to have much credibility.
Scientific findings with high P-values	Science is great as long as the P-values are high.	Quantitative science misses the point. All evaluation must be qualitative.	P-values are a very crude approximation but have a signal. They should ideally be used with care and supplemented with other measures.
Randomized Controlled Trials	If interventions have positive RCTs they are good, and if they don’t, they aren’t.
Quantitative Measurement	Quantitative measurement should be used for all things all the time. An organization should estimate the straightforward things and use that for decision making.	Quantitative measurement always ignores what really matters. We should instead depend on personal intuition on all of these things.
Qualitative Evaluations	Where quantitative measures don’t apply, we can make simple rubrics and have those be used for everything else.
Ontology	Things have Aristotelian Ideals and we try to name things accordingly.	Naming is all useless. It’s a fraud. Nothing really exists.	Names are pragmatic to help people refer to similar things, even if there aren’t external reasons to do so.
Central Planning	A centrally coordinated economy seems ideal, as the government could best control things.	The government can’t be trusted at all. We need anarchy.	The government can do some things okay, but corporations and other groups are better at other things.
Morality	There’s one true morality and it’s fairly simple.	There’s no morality. Everything is relative. We can make no statements that hold about morality.	Morality seems complicated. It seems like there are some principles that are safe. Even when moral principles aren’t sure, we can use reasoning under uncertainty to make safe bets.
Rational Agents	Humans ought to behave like Von Neumann rational agents. Where they don’t the humans are wrong.	There is no such thing as “ought”. It’s all relative.	The Von Neumann model is highly simplistic. There is hypothetically a more complete model that could be used to help humans optimize their best decisions.

Comments from Nuño Sempere

Restored with permission (Nuño’s comments, with Ozzie’s replies).

On “Aristotelian Ideals”:

Nuño Sempere: Platonic Ideals, Aristotelian Categories/Forms

Ozzie Gooen: good point, thanks!

On “There is no such thing as “ought””:

Nuño Sempere: I wouldn’t put Wittgenstein with the anti-rationalists, but he’s the person I associate most strongly with this position.

Ozzie Gooen: He’s a fun one. I think the anti-rationalists have taken inspiration from his ideas and run with them. Also, in some ways, he’s exactly an anti-rationalist in the sense that he made a fairly complete attack on rationalist ideals. He came from the rationalist school of thought, but arguably changed from that.