Swinging Sims

I was gratuitously plugging my Sims paper on Crooked Timber, and I got a few comments about how I didn’t really seem to have got to the heart of the worry Bostrom raises (see here and in the comments thread on my post). At first I was inclined to agree (as in the comments thread to Matthew Yglesias’s post, but on reflection I’ve decided I should have come out swinging harder.

Fodor says somewhere that it’s a law that every technical problem has a technical solution. It looks to some like I’m just raising a technical problem for Bostrom, so it’s a law that it has a solution. If all I’ve done is given Bostrom a reason to read up on some more stats to tidy up the argument, then it’s reasonable to believe that there is a way to tidy up the argument even in advance of his expositing it. So let me try and make the argument as non-technically as possible.

Quick summary of Bostrom’s argument, without the technical details because I want to give the non-technical response. Given some plausible hypotheses about the world, most creatures with experiences kinda like yours are simulations, not material objects. So you’re probably a simulation, not a material object. (The technical objections are all to do with what goes on at the point the arguers say ‘So’. To cut a long story short, there’s an unstated premise here, and it’s inconsistent. Let’s, perhaps generously, bracket those objections.)

We can concede the premise. True, the vast majority of things with experiences kinda like mine are simulated. And what I’m kinda like is part of my evidence. But it isn’t all of it, because I know a lot more than what I’m kinda like. And we should never make judgements from a part of our evidence set when there’s potentially more evidence to use.

To pick a salient example, most things with experiences kinda like mine don’t have their emotional well-being tied to Red Sox wins and losses. No reason yet to conclude that how happy I’ll be this October is more-or-less independent of how the Red Sox do, because I know something more about my particular position in the world.

I actually know quite a lot about my position in the world, on a bit of reflection. There are particular combinations of experiences I’ve had that I can be just about certain no one else in the universe has had or will have. (If the universe is infinite I can’t be so sure, but if the universe is infinite the sense in which ‘most’ creatures like me are simulations becomes elusive. I doubt there’s a higher cardinality of simulated beings than material beings.) What’s the percentage of creatures with experiences just like mine that are simulated, and what’s the percentage that’s material? Bostrom doesn’t know. All he can say is because he doesn’t know, it shouldn’t matter, so I should use the percentages of creatures kinda like me. His motto is ignore evidence when you can’t figure out what effect it has.

What do I say? Well some days I say I do know. I know the percentage of beings with experiences just like mine that are material is 100. In that case, by Bostrom’s reasoning, I should infer I’m material. This isn’t an argument that I’m material, it’s an argument that Bostrom’s reasoning, plus what I know about myself, doesn’t undermine my confidence that I’m material. There’s no reductio argument against knowledge of your own materiality from simulation considerations.

Other days I might waver. At least for the sake of the argument I might, because really I feel I do know the relevant percentage and no one’s given me a good reason to think I don’t. But let’s pretend I’m prepared to doubt this really is knowledge. I still don’t feel any force of Bostrom’s argument. Still, there’s a big chunk of evidence I have that Bostrom says I should ignore, and I don’t know why. He says I should pay close attention to the percentage of people kinda like me that are simulations. I say what about the evidence that tells me more about my place in the universe, evidence that I use every day in reasoning about, for instance, how I’ll feel about the Red Sox. He has to say, “I don’t know how to account for that. Maybe that’s good evidence you’re human, maybe it’s further evidence you’re a simulation, maybe (although this seems implausible) it’s neutral between the two because the percentage of people just like you who are material equals the percentage of people kinda like you who are material. But when we don’t know what to do with evidence we should always take the third option, and treat it like it’s neutral evidence.”

I say that looks like one of the three options, that the evidence is neutral, is being given special status, and I want to know why. (Especially since this is the step that leads to the inconsistencies.) And here I think the simulation argument just goes quiet, because it has nothing more to say. I say (very politely giving myself the last word) that this step needs justification, and without it I’m going to mostly ignore the argument.

If I simply don’t know what effect the largest chunk of my evidence has on the probability that I’m material rather than simulated, then it looks like I simply don’t know the probability that I’m material rather than simulated. And if I don’t know that I might be well within my rights to go on acting as if it’s 1, like I believed all along. Whether that’s within my rights or obligatory depends on some hard questions about the status of epistemic conservatism, and in particular on whether epistemic radicalism is useless or positively misguided. I waver on that question too, so the number of days where I think Bostrom has given me any reason at all to worry is vanishing.

UPDATE: I just noticed that John Turri also sprung to my defence against Yglesias. I think Matthew’s post was quite useful, because it made me think about how to spell out the argument without appeal to technical details. Hence the above rant, which was fun to write at least. (And as Andrew McGonigal quite rightly points out in the comments, I pull some punches here regarding various externalisms, criticisms which again don’t rely on technical details.)

3 Replies to “Swinging Sims”

  1. Does ‘_ is material’ mean in a Sim’s mouth (or head) what it means in a human’s? Surely it should pick out the property that is maximally consistent with the (Sim-English-speaking) Sim community’s use, and the way the world is cut up at the joints. But that seems to me to be unlikely to co-incide or overlap with the property that is maximally consistent with the (English-speaking) human community’s use, and the way the world is cut up at the joints. For example, demonstrations of typical material objects will pick out completely different types of objects in each case, (if computer simulated objects are objects – seems more like ‘fake guns’ then ‘red guns’ to me). So most creatures with experiences kind of like mine are unlikely to be simulations, it seems to me, since the content of their experiences is going to be wholly different. Experiences, like other mental states, are partly individuated by their content, so it seems to me likely that, on at least one fair construal of ‘kinda like’, Sim experiences aren’t at all like mine. (They might be phenomenologically similar, though even that’s contentious, but why should phenomenological similarity trump every other kind?) The creatures with experiences most like mine are likely to be higher order primates, and for the reasons Brian says, I have good grounds for holding that I’m not one of them.

    I’m ignorant of the literature here, although I have skimmed Bostrom’s article. I’m sure this must be a common complaint, and Bostrom must address it somewhere. Any ideas where? What line does he take?

  2. If Bostrom has done a lot on showing how to make his arguments stand up in the face of content externalism I don’t remember it. It seems to me that either externalism about content, or externalism about justification, or externalism about perception would sink the argument, and it seems to me lots of people accept the disjunction of those three externalisms. Rather than repeating the arguments for such views, I was trying here to take on the argument on its own (internalist) terms.

    One word in his defence though. It isn’t clear that simulations don’t inherit the concepts of their creators. Maybe the relevant causal chains in a sims mouth stretch back to what her creator saw, and in that way ‘is material’ denotes a property she lacks. Maybe.

  3. Good, although it’s far from clear that even that latter move would get him off the hook. The original paper motivates the supposed proliferation of Sims via the suggestion that lots of later civilisations might simulate detailed versions of people ‘like their forebears’. But the massive loss of information over time means that it seems wholly unlikely that posthumans can command the kind of individuating thought about e.g. the particulars that surround me to simulate them effectively – even their singular demonstrative concepts won’t count as being about e.g. the table in front of me, rather than some generic table. But my experience, memory, etc., constitutively includes such singular thought about the particulars in my environment. So it seems that, even if the Sims inherit the posthuman concepts of their creators, their experience might not be noticeably similar to mine, unless we give undue prominence to (internally individuated) phenomenological similarity.

    There’s also the question of whether post-humans can genuinely share my conceptual range – can we, as we now are, think about e.g. witchcraft, or the moon, in the same way that primitive humans did? A moderate holism about concept-individuation seems rather plausible to me. Maybe the vastly advanced beliefs of post-humans mean that there’s aspects of contemporary human thought that they, (and therefore Sims, on the inheritance model) simply couldn’t share. Again, since I employ such concepts, the Sims’ experience may end up significantly unlike mine.

    In any case, the post wasn’t intended to speak directly to Brian’s strategy: I agree that it’s more satisfying to beat them at their own game.

Comments are closed.