abramdemski

abramdemski's Posts

Sorted by New

abramdemski's Comments

New Top EA Causes for 2020?

Bayesian Reformation

Bayesian orthodoxy has often been compared to a cult. Here, we take the opposite perspective, comparing Bayesianism to a long-standing institution like the Catholic Church.

The protestant reformation created initial (and ongoing) unrest, including bloody violence. However, we take the view that the reformation improved social institutions generally, culminating in the industrial revolution. Hence, the reformation offers a model of high-leverage changes with dramatically positive consequences on the future.

The primary question, then, is effective instigation and agitation -- how does one create conditions ripe for Reformation? What are the preconditions to a modern Martin Luther, nailing a new 95 Theses to the door of Bayesianism and Effective Altruism?

Since the historical Martin Luther initially studied as a lawyer, but hated it, we recommend a critical study of the law as a starting point, including the naturalization of the social contract. This goes well with a general study of the failures of top-down optimization such as Bayesian/utilitarian public policy tends to favor.

The distribution of radical literature being an old standby of the instigation and agitation industry, we recommend the circulation of literature concerning radical probabilism, a doctrine which critiques the Bayesian update -- insisting that, while Bayes' Law correctly analyzes updating on a certainty, there is in fact no such thing as certain knowledge. Bayes' Law therefore never applies.

According to this doctrine, all we can say about the shifting nature of belief is that they should not be (unboundedly) Dutch Book -able. In particular, like financial markets, they should behave like a martingale -- at any time, the current betting value ('price') of any belief should also be a best estimate of its future value.

We also broadly recommend attention be paid to critiques of Bayesian orthodoxy, implying that we sit down and actually pay attention to various critiques of Bayesianism (Frequentist, and otherwise). For too long have we rested in complacency, accepting Bayesian dogma. The time has come to develop a better perspective.