Eliezer, where do your strong claims about the causal structure of scientific discourse come from?
I consider them as obvious first-order approximations, especially to the normative structure. Does an authoritative expert cause a hypothesis to become true, so that we can surgically intervene on the truth of a physical theory by giving its adherents more authority? Clearly not. Does an authoritative expert cause the “arguments” to become stronger? Defining the matter normatively makes it clear that the answer is no. If we talk about perceived arguments, then a good expert makes us perceive the arguments as stronger, but that’s simply a question of backward inference not causation—like saying that if the sidewalk is slippery this causes us to think it is raining, but does not cause it to rain.
Since I am discussing what we should pay attention to, not what we do pay attention to, it makes sense to discuss the normative causal struture.
Do you have an alternative suggestion? Clearly there are many things that supervene on expert opinion besides valid arguments, which we could coalesce into a Noise node and a Bias node, describing the invalid influences that we think we can’t predict and that we think we can systematically predict respectively:
Truth → Argument → Expert Belief ← Noise, Bias
This gives us obvious inferences like “If you know the experts will be biased, but you don’t understand their arguments apart from authority, you will be less certain of the truth” and “Surgical interventions on bias and on expert belief cannot make a proposition true, or change which non-authoritative propositions are arguments in favor of it”.
You probably have that directional causal structure represented in your mind, which makes the above inferences seem plausible; I just wrote it out.
Eliezer, where do your strong claims about the causal structure of scientific discourse come from?
I consider them as obvious first-order approximations, especially to the normative structure. Does an authoritative expert cause a hypothesis to become true, so that we can surgically intervene on the truth of a physical theory by giving its adherents more authority? Clearly not. Does an authoritative expert cause the “arguments” to become stronger? Defining the matter normatively makes it clear that the answer is no. If we talk about perceived arguments, then a good expert makes us perceive the arguments as stronger, but that’s simply a question of backward inference not causation—like saying that if the sidewalk is slippery this causes us to think it is raining, but does not cause it to rain.
Since I am discussing what we should pay attention to, not what we do pay attention to, it makes sense to discuss the normative causal struture.
Do you have an alternative suggestion? Clearly there are many things that supervene on expert opinion besides valid arguments, which we could coalesce into a Noise node and a Bias node, describing the invalid influences that we think we can’t predict and that we think we can systematically predict respectively:
Truth → Argument → Expert Belief ← Noise, Bias
This gives us obvious inferences like “If you know the experts will be biased, but you don’t understand their arguments apart from authority, you will be less certain of the truth” and “Surgical interventions on bias and on expert belief cannot make a proposition true, or change which non-authoritative propositions are arguments in favor of it”.
You probably have that directional causal structure represented in your mind, which makes the above inferences seem plausible; I just wrote it out.