I think you are right that this might be a norm/heuristic in the community, but in the spirit of a “justificatory story of our epistemic practices,” I want to look a little more at
4. When arguments lead us to conclusions that are both speculative and fanatical, treat this as a sign that something has gone wrong.
First, I’m not sure that “speculative” is an independent reason that conclusions are discounted, in the sense of a filter that is applied ex-post. In your 15AI thought experiment, for example, I think that expected value calculations would get you most of the way toward explaining an increase in fanaticism; the probability that we can solve the problem might increase on net, despite the considerations you note about replication. The remaining intuition might be explained by availability/salience bias, to which EA is not immune.
Now, “speculative” scenarios might be discounted during the reasoning process if we are anchored to commonsense priors, but this would fall under typical bayesian reasoning. The priors we use and the weight we grant various pieces of evidence are still epistemic norms worthy of examination! But a different kind than suggested by the fourth principle.
Suppose “speculative” arguments are discounted ex-post in EA. I think this practice can still be redeemed on purely bayesian grounds as a correction to the following problems:
Undiscovered Evidence: An argument seems speculative not just insofar as it is divorced from empirical observations, but also insofar as we have not thought about it very much. It seems that AI risk has become less speculative as people spend more time thinking about it, holding constant actual progress in AI capabilities. We have some sense of the space of possible arguments that might be made and evidence that might be uncovered, given further research on a topic. And these undiscovered arguments/evidence might not enter neatly into our initial reasoning process. We want some way to say “I haven’t thought of it yet, but I bet there’s a good reason this is wrong,” as we might respond to some clever conspiracy theorist who presents a superficially bulletproof case for a crazy theory we haven’t encountered before. And discounting speculative conclusions is one way to achieve this.
This point is especially relevant for speculative conclusions because they often rely on chains of uncertain premises, making our credence in their conclusions all the more sensitive to new information that could update multiple steps of the argument.
Model Uncertainty: Even in a domain where we have excavated all the major arguments available to us, we may still suffer from “reasoning in the dark,” ie, in the absence of solid empirics. When reasoning about extremely unlikely events, the probability our model is wrong can swamp our credence in its conclusion. Discounting speculative conclusions allows us to say “we should be fanatical insofar as my reasoning is correct, but I am not confident in my reasoning.”
We can lump uncertainty in our axiology, epistemology, and decision theory under this section. That is, a speculative conclusion might look good only under total utilitarian axiology, bayesian epistemology, and causal decision theory, but a more conventional conclusion might be more robust to alternatives in these categories. (Note that this is a prior question to the evidential-hedging double bind set up in Appendix B.)
Chains of uncertain premises also make model uncertainty doubly important for speculative conclusions. As Anders Sandberg points out, “if you have a long argument, the probability of there being some slight error somewhere is almost 1.”
Even after accounting for these considerations, we might find that the EV of pursuing the speculative path warrants fanaticism. In this event, discounting the speculative conclusion might be a pragmatic move to deprioritize actions on this front in anticipation of new evidence that will come to light, including evidence that will bear on model uncertainty. (We might treat this as a motivation for imprecise credences, prioritizing views with sharper credences over speculative views with fuzzier ones.)
I understand you to be offering two potential stories to justify ‘speculativeness-discounting’.
First, EAs don’t (by and large) apply a speculativeness-discount ex post. Instead, there’s a more straightforward ‘Bayesian+EUM’ rationalization of the practice. For instance, the epistemic practice of EAs may be better explained with reference to more common-sense priors, potentially mediated by orthodox biases.
Or perhaps EAs do apply a speculativeness-discount ex post. This too can be justified on Bayesian grounds.
We often face doubts about our ability to reason through all the relevant considerations, particularly in speculative domains. For this reason, we update on higher-order uncertainty, and implement heuristics which themselves are justified on Bayesian grounds.
In my response, I’ll assume that your attempted rationale for Principle 4 involves justifying the norm with respect to the following two views:
Expected Utility Maximization (EUM) is the optimal decision-procedure.
The relevant probabilities to be used as inputs into our EUM calculation are our subjective credences.
The ‘Common Sense Priors’ Story
I think your argument in (1) is very unlikely to provide a rationalization of EA practice on ‘Bayesian + EUM’ grounds.[1]
Take Pascal’s Mugging. The stakes can be made high enough that the value involved can easily swamp your common-sense priors. Of course, people have stories for why they shouldn’t give the money to the mugger. But these stories are usually generated because handing over their wallet is judged to be ridiculous, rather than the judgment arising from an independent EU calculation. I think other fanatical cases will be similar. The stakes involved under (e.g.) various religious theories and our ability to acausally affect an infinite amount of value are simply going to be large enough to swamp our initial common-sense priors.
Thus, I think the only feasible ‘Bayes+EUM’ justification you could offer would have to rely on your ‘higher-order evidence’ story about the fallibility of our first-order reasoning, which we’ll turn to below.
The ‘Higher-Order Evidence’ Story
I agree that we can say: “we should be fanatical insofar as my reasoning is correct, but I am not confident in my reasoning.”
The question, then, is how to update after reflecting on your higher-order evidence. I can see two options: either you have some faith in your first-order reasoning, or no faith.
Let’s start with the case where you have some faith in your first-order reasoning. Higher-order evidence about your own reasoning might decrease the confidence in your initial conclusion. But, as you note, “we might find that the EV of pursuing the speculative path warrants fanaticism”. So, what to do in that case?
I think it’s true that many people will cite considerations of the form “let’s pragmatically deprioritize the high EV actions that are both speculative and fanatical, in anticipation of new evidence”. I don’t think that provides a sufficient justificatory story of the epistemic norms to which most of us hold ourselves.
Suppose we knew that our evidential situation was as good as it’s ever going to be. Whatever evidence we currently have about (e.g.) paradoxes in infinite ethics, or the truth of various religions constitutes ~all the evidence we’re ever going to have.
I still don’t expect people to follow through on the highest EV option, when that option is both speculative and fanatical.
Or maybe you have a bounded utility function. In that case, imagine that the world already contains a sufficiently large number of suffering entities. How blase are you, really, about the creation of arbitrarily many suffering-filled hellscapes?
There’s more to say here, but the long and short of it is: if you fail to reach a point where you entirely discount certain forms of speculative reasoning, I don’t think you’ll be able to recover anything like Principle 4. My honest view is that many EAs have a vague hope that such theories will recover something approaching normality, but very few people actually try to trace out the implications of such theories on their own terms, and follow through on these implications. I’m sympathetic to this quote from Paul Christiano:
I tried to answer questions like “How valuable is it to accelerate technological progress?” or “How bad is it if unaligned AI takes over the world?” and immediately found that EU maximization with anything like “utility linear in population size” seemed to be unworkable in practice. I could find no sort of common-sensical regularization that let me get coherent answers out of these theories, and I’m not sure what it would look like in practice to try to use them to guide our actions.
Higher-Order Evidence and Epistemic Learned Helplessness
Maybe you’d like to say: “in certain domains, we should assign our first-order calculations about which actions maximize EU zero weight. The heuristic ‘sometimes assign first-order reasoning zero weight’ can be justified on Bayesian grounds.”
I agree that we should sometimes assign our first-order calculations about which actions maximize EU zero weight. I’m doubtful that Bayesianism or EUM play much of a role in explaining why this norm is justified.
When we’re confronted with the output of an EUM calculation that feels off, we should listen to the parts of us which tell us to check again, and ask why we feel tempted to check again.
If we’re saying “no, sorry, sometimes I’m going to put zero weight on a subjective EU calculation”, then we’re already committed to a view under which subjective EU calculations only provide action-guidance in the presence of certain background conditions.
If we’re willing to grant that, then I think the interesting justificatory story is a story which informs us of what the background conditions for trusting EU calculations actually are — rather than attempts to tell post hoc stories about how our practices can ultimately be squared with more foundational theories like Bayesianism + EUM.
If you’re interested, I’ll have a post in April touching on these themes. :)
I also think the sociological claim you made is probably false. However, as you’re primarily asking about the justificatory side of things, I’ll bracket that here — though I’m happy to make this case in more detail if you’d like.
Thanks for the thorough response! I agree with a lot of what you wrote, especially the third section on Epistemic Learned Helplessness: “Bayesianism + EUM, but only when I feel like it” is not a justification in any meaningful sense.
On Priors
I agree that we can construct thought experiments (Pascal’s Mugging, acausal trade) with arbitrarily high stakes to swamp commonsense priors (even without religious scenarios or infinite value, which are so contested I think it would be difficult to extract a sociological lesson from them).
On Higher Order Evidence
I still think a lot of speculative conclusions we encounter in the wild suffer from undiscovered evidence and model uncertainty, and even barring this we might want to defer taking action until we’ve had a chance to learn more.
Your response jumps over these cases to those where we have “~all the evidence we’re ever going to have,” but I’m skeptical these cases exist. Even with religion, we might expect some future miracles or divine revelations to provide new evidence; we have some impossibility theorems in ethics, but new ideas might come to light that resolve paradoxes or avoid them completely. In fact, soteriological research and finding the worldview that best acausally benefits observers are proposals to find new evidence.
But ok, yes, I think we can probably come up with cases where we do have ~all the evidence and still refrain from acting on speculative + fanatical conclusions.
Problem 1: Nicheness
From here on, I’m abandoning the justification thing. I agree that we’ve found some instances where the Fourth Principle holds without Bayesian + EUM justification. Instead, I’m getting more into the semantics of what is a “norm.”
The problem is that the support for this behavior among EAs comes from niche pieces of philosophy like Pascal’s Mugging, noncausal decision theory, and infinite ethics, ideas that are niche not just relative to the general population, but also within EA. So I feel like the Fourth Principle amounts to “the minority of EAs who are aware of these edge cases behave this way when confronted with them,” which doesn’t really seem like a norm about EA.
Problem 2: Everyone’s Doing It
(This is also not a justification, it’s an observation about the Fourth Principle)
The first three principles capture ways that EA differs from other communities. The Fourth Principle, on the other hand, seems like the kind of thing that all people do? For example, a lot of people write off earning to give when they first learn about it because it looks speculative and fanatical. Now, maybe EAs differ from other people on which crazy train stop they deem “speculative,” and I think that would qualify as a norm, but relative to each person’s threshold for “speculative,” I think this is more of a human-norm than an EA-norm.
Would love your thoughts on this, and I’m looking forward to your April post :)
Thanks for the excellent post!
I think you are right that this might be a norm/heuristic in the community, but in the spirit of a “justificatory story of our epistemic practices,” I want to look a little more at
First, I’m not sure that “speculative” is an independent reason that conclusions are discounted, in the sense of a filter that is applied ex-post. In your 15AI thought experiment, for example, I think that expected value calculations would get you most of the way toward explaining an increase in fanaticism; the probability that we can solve the problem might increase on net, despite the considerations you note about replication. The remaining intuition might be explained by availability/salience bias, to which EA is not immune.
Now, “speculative” scenarios might be discounted during the reasoning process if we are anchored to commonsense priors, but this would fall under typical bayesian reasoning. The priors we use and the weight we grant various pieces of evidence are still epistemic norms worthy of examination! But a different kind than suggested by the fourth principle.
Suppose “speculative” arguments are discounted ex-post in EA. I think this practice can still be redeemed on purely bayesian grounds as a correction to the following problems:
Undiscovered Evidence: An argument seems speculative not just insofar as it is divorced from empirical observations, but also insofar as we have not thought about it very much. It seems that AI risk has become less speculative as people spend more time thinking about it, holding constant actual progress in AI capabilities. We have some sense of the space of possible arguments that might be made and evidence that might be uncovered, given further research on a topic. And these undiscovered arguments/evidence might not enter neatly into our initial reasoning process. We want some way to say “I haven’t thought of it yet, but I bet there’s a good reason this is wrong,” as we might respond to some clever conspiracy theorist who presents a superficially bulletproof case for a crazy theory we haven’t encountered before. And discounting speculative conclusions is one way to achieve this.
This point is especially relevant for speculative conclusions because they often rely on chains of uncertain premises, making our credence in their conclusions all the more sensitive to new information that could update multiple steps of the argument.
Model Uncertainty: Even in a domain where we have excavated all the major arguments available to us, we may still suffer from “reasoning in the dark,” ie, in the absence of solid empirics. When reasoning about extremely unlikely events, the probability our model is wrong can swamp our credence in its conclusion. Discounting speculative conclusions allows us to say “we should be fanatical insofar as my reasoning is correct, but I am not confident in my reasoning.”
We can lump uncertainty in our axiology, epistemology, and decision theory under this section. That is, a speculative conclusion might look good only under total utilitarian axiology, bayesian epistemology, and causal decision theory, but a more conventional conclusion might be more robust to alternatives in these categories. (Note that this is a prior question to the evidential-hedging double bind set up in Appendix B.)
Chains of uncertain premises also make model uncertainty doubly important for speculative conclusions. As Anders Sandberg points out, “if you have a long argument, the probability of there being some slight error somewhere is almost 1.”
Even after accounting for these considerations, we might find that the EV of pursuing the speculative path warrants fanaticism. In this event, discounting the speculative conclusion might be a pragmatic move to deprioritize actions on this front in anticipation of new evidence that will come to light, including evidence that will bear on model uncertainty. (We might treat this as a motivation for imprecise credences, prioritizing views with sharper credences over speculative views with fuzzier ones.)
Thanks for the comment!
(Fair warning, my response will be quite long)
I understand you to be offering two potential stories to justify ‘speculativeness-discounting’.
First, EAs don’t (by and large) apply a speculativeness-discount ex post. Instead, there’s a more straightforward ‘Bayesian+EUM’ rationalization of the practice. For instance, the epistemic practice of EAs may be better explained with reference to more common-sense priors, potentially mediated by orthodox biases.
Or perhaps EAs do apply a speculativeness-discount ex post. This too can be justified on Bayesian grounds.
We often face doubts about our ability to reason through all the relevant considerations, particularly in speculative domains. For this reason, we update on higher-order uncertainty, and implement heuristics which themselves are justified on Bayesian grounds.
In my response, I’ll assume that your attempted rationale for Principle 4 involves justifying the norm with respect to the following two views:
Expected Utility Maximization (EUM) is the optimal decision-procedure.
The relevant probabilities to be used as inputs into our EUM calculation are our subjective credences.
The ‘Common Sense Priors’ Story
I think your argument in (1) is very unlikely to provide a rationalization of EA practice on ‘Bayesian + EUM’ grounds.[1]
Take Pascal’s Mugging. The stakes can be made high enough that the value involved can easily swamp your common-sense priors. Of course, people have stories for why they shouldn’t give the money to the mugger. But these stories are usually generated because handing over their wallet is judged to be ridiculous, rather than the judgment arising from an independent EU calculation. I think other fanatical cases will be similar. The stakes involved under (e.g.) various religious theories and our ability to acausally affect an infinite amount of value are simply going to be large enough to swamp our initial common-sense priors.
Thus, I think the only feasible ‘Bayes+EUM’ justification you could offer would have to rely on your ‘higher-order evidence’ story about the fallibility of our first-order reasoning, which we’ll turn to below.
The ‘Higher-Order Evidence’ Story
I agree that we can say: “we should be fanatical insofar as my reasoning is correct, but I am not confident in my reasoning.”
The question, then, is how to update after reflecting on your higher-order evidence. I can see two options: either you have some faith in your first-order reasoning, or no faith.
Let’s start with the case where you have some faith in your first-order reasoning. Higher-order evidence about your own reasoning might decrease the confidence in your initial conclusion. But, as you note, “we might find that the EV of pursuing the speculative path warrants fanaticism”. So, what to do in that case?
I think it’s true that many people will cite considerations of the form “let’s pragmatically deprioritize the high EV actions that are both speculative and fanatical, in anticipation of new evidence”. I don’t think that provides a sufficient justificatory story of the epistemic norms to which most of us hold ourselves.
Suppose we knew that our evidential situation was as good as it’s ever going to be. Whatever evidence we currently have about (e.g.) paradoxes in infinite ethics, or the truth of various religions constitutes ~all the evidence we’re ever going to have.
I still don’t expect people to follow through on the highest EV option, when that option is both speculative and fanatical.
Under MEC, EAs should plausibly be funneling all their money into soteriological research. Or perhaps you don’t like MEC, and think we should work out the most plausible worldview under which we can affect strongly-Ramsey-many sentient observers.[2]
Or maybe you have a bounded utility function. In that case, imagine that the world already contains a sufficiently large number of suffering entities. How blase are you, really, about the creation of arbitrarily many suffering-filled hellscapes?
There’s more to say here, but the long and short of it is: if you fail to reach a point where you entirely discount certain forms of speculative reasoning, I don’t think you’ll be able to recover anything like Principle 4. My honest view is that many EAs have a vague hope that such theories will recover something approaching normality, but very few people actually try to trace out the implications of such theories on their own terms, and follow through on these implications. I’m sympathetic to this quote from Paul Christiano:
Higher-Order Evidence and Epistemic Learned Helplessness
Maybe you’d like to say: “in certain domains, we should assign our first-order calculations about which actions maximize EU zero weight. The heuristic ‘sometimes assign first-order reasoning zero weight’ can be justified on Bayesian grounds.”
I agree that we should sometimes assign our first-order calculations about which actions maximize EU zero weight. I’m doubtful that Bayesianism or EUM play much of a role in explaining why this norm is justified.
When we’re confronted with the output of an EUM calculation that feels off, we should listen to the parts of us which tell us to check again, and ask why we feel tempted to check again.
If we’re saying “no, sorry, sometimes I’m going to put zero weight on a subjective EU calculation”, then we’re already committed to a view under which subjective EU calculations only provide action-guidance in the presence of certain background conditions.
If we’re willing to grant that, then I think the interesting justificatory story is a story which informs us of what the background conditions for trusting EU calculations actually are — rather than attempts to tell post hoc stories about how our practices can ultimately be squared with more foundational theories like Bayesianism + EUM.
If you’re interested, I’ll have a post in April touching on these themes. :)
I also think the sociological claim you made is probably false. However, as you’re primarily asking about the justificatory side of things, I’ll bracket that here — though I’m happy to make this case in more detail if you’d like.
Presumably acausally.
Thanks for the thorough response! I agree with a lot of what you wrote, especially the third section on Epistemic Learned Helplessness: “Bayesianism + EUM, but only when I feel like it” is not a justification in any meaningful sense.
On Priors
I agree that we can construct thought experiments (Pascal’s Mugging, acausal trade) with arbitrarily high stakes to swamp commonsense priors (even without religious scenarios or infinite value, which are so contested I think it would be difficult to extract a sociological lesson from them).
On Higher Order Evidence
I still think a lot of speculative conclusions we encounter in the wild suffer from undiscovered evidence and model uncertainty, and even barring this we might want to defer taking action until we’ve had a chance to learn more.
Your response jumps over these cases to those where we have “~all the evidence we’re ever going to have,” but I’m skeptical these cases exist. Even with religion, we might expect some future miracles or divine revelations to provide new evidence; we have some impossibility theorems in ethics, but new ideas might come to light that resolve paradoxes or avoid them completely. In fact, soteriological research and finding the worldview that best acausally benefits observers are proposals to find new evidence.
But ok, yes, I think we can probably come up with cases where we do have ~all the evidence and still refrain from acting on speculative + fanatical conclusions.
Problem 1: Nicheness
From here on, I’m abandoning the justification thing. I agree that we’ve found some instances where the Fourth Principle holds without Bayesian + EUM justification. Instead, I’m getting more into the semantics of what is a “norm.”
The problem is that the support for this behavior among EAs comes from niche pieces of philosophy like Pascal’s Mugging, noncausal decision theory, and infinite ethics, ideas that are niche not just relative to the general population, but also within EA. So I feel like the Fourth Principle amounts to “the minority of EAs who are aware of these edge cases behave this way when confronted with them,” which doesn’t really seem like a norm about EA.
Problem 2: Everyone’s Doing It
(This is also not a justification, it’s an observation about the Fourth Principle)
The first three principles capture ways that EA differs from other communities. The Fourth Principle, on the other hand, seems like the kind of thing that all people do? For example, a lot of people write off earning to give when they first learn about it because it looks speculative and fanatical. Now, maybe EAs differ from other people on which crazy train stop they deem “speculative,” and I think that would qualify as a norm, but relative to each person’s threshold for “speculative,” I think this is more of a human-norm than an EA-norm.
Would love your thoughts on this, and I’m looking forward to your April post :)