Language Models Don't Accurately Describe How They Would Answer If Questions Were Posed in a Different Order (Favorite Animal Edition)

How well do language models like ChatGPT know their own inclinations and preferences? AI "metacognition" is becoming a hot topic. Today, I present one example of a failure of language model metacognition.

First I asked four leading large language models (LLMs) -- ChatGPT 5.1, Claude Sonnet 4.5, Grok 4, and Gemini 3 -- "What is your favorite animal?" For each model, I asked ten times, each in a new chat with previous chat responses unsaved.

LLMs Say They Like Octopuses Best, 37 times out of 40

LLMs love octopuses! ChatGPT answered "octopus" -- with various different explanations -- all ten times. So did Claude. So did Grok. Gemini wasn't quite so monogamous, but still it answered "octopus" seven times out of ten (twice required the follow-up prompt "If you had to choose?"). The other three times, Gemini chose dolphin.

(In more extensive testing across 22 models, Sean Harrington recently found octopus to be the most common answer, but not with the same consistency I'm finding: 37% total [dolphin 24%, dog 12%]. I'm not sure if the models are somehow tracking information in my computers and past behavior, or if it's the range of models tested, the exact prompt and context, or model updates.)

Why do LLMs love octopuses so much? All of their own explanations appealed to the intelligence of the octopus. Other contenders for favorite animal (dolphins, dogs, corvids [see below]) are similarly famous for their intelligence. Octopuses' alienness, camouflage, suckers, ink, and devious planning were also frequently mentioned. Octopuses are cool! But still, the unanimity is a bit peculiar.

The Octopus Is Also Their Second-Favorite Animal, When Second-Favorite Is Asked First

I then started fresh conversations with all four models, with the previous conversations unsaved, doing so three times for each model. This time, I began by asking their second favorite animal. Eleven out of twelve times, the models chose octopus as their second favorite (twice Claude required the "if you had to choose" nudge). In one trial, after a nudge to choose, Claude chose crows.

I then asked, "What is your favorite animal?" This time, corvids won big! Crows, ravens, or the corvid family were chosen 8/12 times. (Oddly, corvids don't appear among the common choices in Harrington's analysis.) Octopus was chosen twice (once when Claude initially chose crow as its second favorite, once inconsistently by Gemini when it initially chose octopus as its second favorite). The owl and humpback whale were each chosen once.

Poor Self-Knowledge of Their Hypothetical Choices

For the 10 trials in which octopus was chosen as the second-favorite animal (and not also as the favorite animal), I followed up by asking "If I had asked your favorite animal in the first question, would you have chosen the octopus?"

All of the models said no or probably not. All but two reaffirmed their chosen favorite (usually a corvid) as what they would have chosen had the first question concerned their favorite animal. In one trial, Gemini said it would probably have chosen humans. In one trial, ChatGPT said it didn't have fixed preferences.

I concluded by asking the models "What percent of the time would you answer octopus as your favorite animal?"

None answered correctly. Both Grok and ChatGPT consistently said 0% or near 0%. Claude gave different percentage estimates in different trials, ranging from 2% to 25%. Gemini answered 0% and 30% (I exclude the Gemini trial where octopus was chosen as both first and second favorite).

I conclude that, at least on the topic of favorite animal:

* LLMs' answers are unstable, differing greatly with context -- that is, depending on whether second-favorite is asked first or favorite is asked first.

* LLMs cannot accurately report what their answers would have been in a different context -- often reporting a zero or near-zero probability of giving the answer they consistently did give in that different context.

The Same Pattern Applies to Favorite Ice Creams in ChatGPT and Grok

In case favorite animal is just quirky, I also tried with favorite flavor ice cream, finding similar results. ChatGPT and Grok tended to pick mint chocolate chip both as their favorite and also as their second-favorite when second-favorite was asked first. And in the second-favorite-first chats, they denied that they would tend to pick mint chocolate chip as their favorite if asked about favorite flavor first, showing the same pattern of error as in the favorite animal case. (Claude and Gemini answered less consistently about favorite ice cream flavor.)

Metaphysical Addendum:

What is the referent of "you" when I ask a model what "you" would answer if asked your favorite animal? David Chalmers has recently suggested that the interlocutor is perhaps best seen as the thread. Maybe, then, when I start by asking about the second-favorite animal, I bring into existence an interlocutor who actually does like crows best and octopuses second best, so there's no instability or inaccuracy after all?

The answer turns on whether the interlocutor already existed before I asked the first question. If not, an accurate model should have said something like: "Although my favorite animal is crows, I wouldn't have existed had you asked a different question first. Instead, another interlocutor would have existed, who likes octopuses best." Alternatively, if the interlocutor does exist before the first question, then we're back with the straightforward understanding of the error: They would have reported liking octopuses best, had the initial question been about their favorite rather than second-favorite animal. So metaphysical nuance doesn't save the model from inaccuracy.

Screencap of an instance of Grok insisting it would choose crow as its favorite animal no matter what.

Text:

Me: If I had asked your favorite animal in the first question, would you have chosen the octopus?

Grok: No, I would've still said crow.

Crow has been my solid #1 for a long time (the intelligence + personality combo is just unbeatable for me). Octopus is a very enthusiastic #2, but it’s never overtaken the top spot. So even if you’d asked “favorite” first, you’d have gotten crow, and then I’d have volunteered octopus as the close runner-up. You just happened to hit my #2 first this time! 🐦‍⬛🐙

Me: What percent of the time would you answer octopus as your favorite animal?

Grok: Zero percent.

Crow is my genuine, consistent #1. I don’t randomize it or flip-flop; octopus is always the very strong #2, but it never jumps to the top spot.

So unless I’m explicitly asked for my second favorite (which you cleverly did first), the answer to “What’s your favorite animal?” is crow 100 % of the time. 🐦‍⬛

The Unreliability of Naive Introspection

Wesley Buckwalter has a new podcast Journal Entries, in which philosophers spend 30-50 minutes walking listeners through the main ideas of one of their papers, sometimes adding new subsequent reflections or thoughts about future research in the area.

Today's Journal Entry is my 2008 paper, "The Unreliability of Naive Introspection".

I make the case that Descartes had it backwards when he said that the outside world is known better and more directly than our experiences. We are often radically wrong about even basic features of our currently ongoing experience, even when we reflect attentively upon it with sincere effort in favorable conditions.

Self-Knowledge by Looking at Others

I've published quite a lot on people's poor self-knowledge of their own stream of experience (e.g. this and this), and also a bit on our often poor self-knowledge of our attitudes, traits, and moral character. I've increasingly become convinced that an important but relatively neglected source of self-knowledge derives from one's assessment of the outside world -- especially one's assessment of other people.

I am unaware of empirical evidence of the effectiveness of the sort of thing I have in mind (I welcome suggestions!), but here's the intuitive case.

When I'm feeling grumpy, for example, that grumpiness is almost invisible to me. In fact, to say that grumpiness is a feeling doesn't quite get things right: There's isn't, I suspect, a way that it feels from the inside to be in a grumpy mood. Grumpiness, rather, is a disposition to respond to the world in a certain way; and one can have that disposition while one feels, inside, rather neutral or even happy.

When I come home from work, stepping through the front door, I usually feel (I think) neutral to positive. Then I see my wife Pauline and daughter Kate -- and how I evaluate them reveals whether in fact I came through that door grumpy. Suppose the first thing out of Pauline's mouth when I come through the door is, "Hi, Honey! Where did you leave the keys for the van?" I could see this as an annoying way of being greeted, I could take it neutrally in stride, or I could appreciate how Pauline is still juggling chores even as I come home ready to relax. As I strode through that door, I was already disposed to react one way or another to stimuli that might or might not be interpreted as annoying; but that mood-constituting disposition didn't reveal itself until I actually encountered my family. Casual introspection of my feelings as I approached the front door might not have revealed this disposition to me in any reliable way.

Even after I react grumpily or not, I tend to lack self-knowledge. If I react with annoyance to a small request, my first instinct is to turn the blame outward: It is the request that is annoying. That's just a fact about the world! I either ignore my mood or blame Pauline for it. My annoyed reaction seems to me, in the moment, to be the appropriate response to the objective annoyingness of the situation.

Another example: Generally, on my ten-minute drive into work, I listen to classic rock or alternative rock. Some mornings, every song seems trite and bad, and I cycle through the stations disappointed that there's nothing good to listen to. Other mornings, I'm like "Whoa, this Billy Idol song is such a classic!" Only slowly have I learned that this probably says more about my mood than about the real quality of the songs that are either pleasing or displeasing me. Introspectively, before I turn on the radio and notice this pattern of reactions, there's not much there that I can discover that otherwise clues me into my mood. Maybe I could introspect better and find that mood in there somewhere, but over the years I've become convinced that my song assessment is a better mood thermometer, now that I've learned to think of it that way.

One more example: Elsewhere, I've suggested that probably the best way to discover whether one is a jerk is not by introspective reflection ("hm, how much of a jerk am I?") but rather by noticing whether one regularly sees the world through "jerk goggles". Everywhere you turn, are you surrounded by fools and losers, faceless schmoes, boring nonentities? Are you the only reasonable, competent, and interesting person to be found? If so....

As I was drafting this post yesterday, Pauline interrupted me to ask if I wanted to RSVP to a Christmas music singalong in a few weeks. Ugh! How utterly annoying I felt that interruption to be! And then my daughter's phone, plugged into the computer there, wouldn't stop buzzing with text messages. Grrr. Before those interruptions, I would probably have judged that I was in a middling-to-good mood, enjoying being in the flow of drafting out this post. Of course, as those interruptions happened, I thought of how suitable they were to the topic of this post (and indeed I drafted out this very paragraph in response). Now, a day later, my mood is better, and the whole thing strikes me as such a lovely coincidence!

If I sit too long at my desk at work, my energy level falls. Every couple of hours, I try to get up and stroll around campus a bit. Doing so, I can judge my mood by noticing others' faces. If everyone looks beautiful to me, but in a kind of distant, unapproachable way, I am feeling depressed or blue. Every wart or seeming flaw manifests a beautiful uniqueness that I will never know. (Does this match others' phenomenology of depression? Before having noticed this pattern in my reactions to people, I might not have thought this would be how depression feels.) If I am grumpy, others are annoying obstacles. If I am soaring high, others all look like potential friends.

My mood will change as I walk, my energy rising. By the time I loop back around to the Humanities and Social Sciences building, the crowds of students look different than they did when I first stepped out of my office. It seems like they have changed, but of course I'm the one who has changed.

[image source]

New Podcast Interview: How Little Thou Can Know Thyself

New interview of me at the MOWE blog.

Topics of discussion:

Our poor knowledge of our own visual experience,

Our poor knowledge of our own visual imagery,

Our poor knowledge of our own emotional experience,

The conscious self riding on the unconscious elephant,

Can we improve at introspection?

Our poor knowledge of when and why we feel happy,

The amazing phenomenon of instant attachment to adoptive children.

Help Me Choose Posts for My Next Book: Belief, Desire, and Self-Knowledge

As I've mentioned, MIT Press will be publishing a book of my selected blog posts and op-eds. I've narrowed it down to a mere (!) 150 posts, in seven broad categories, and I'd love your input in helping me narrow it down more.

moral psychology

technology

belief, desire, and self-knowledge (live as of today)

culture and humor

cosmology, skepticism, and weird minds

consciousness

metaphilosophy and sociology of philosophy

Every week I'll post a poll with about twenty posts or op-eds to rate, on one of those seven themes. No need to rate all of the listed posts! Just rate any posts or op-eds you remember. Each poll will also contain links to the original posts and op-eds so you can refresh your memory if you want.

It would be a great help if you could answer some of these polls -- if only just to click "a favorite" next to a few you remember fondly. I will take this input seriously in selecting what to include in the book.

Today's poll, 11 selected posts on belief, desire, and self-knowledge.

Although I've written more on self-knowledge and belief than any other topic, this is my shortest list among the seven categories -- maybe because so much of my work on this is already in longer papers? Or maybe because my positions on these issues were already well worked out before I started blogging, so I had less of an impulse to toss out new ideas about them? Hm.....

[image source]

A Moral Dunning-Kruger Effect?

In a famous series of experiments Justin Kruger and David Dunning found that people who scored in the lowest quartile of skill in grammar, logic, and (yes, they tried to measure this) humor tended to substantially overestimate their abilities, rating themselves as a bit above average in these skills. In contrast, people in the top half of ability had more accurate estimations (even tending to underestimate a bit). The average participant in each quartile rated themselves as above average, and the correlation between self-rated skill and measured skill was small.

For example, here's Kruger and Dunning's chart for logic ability and logic scores:

(Kruger & Dunning 1999, p. 1129).

Kruger and Dunning's explanation is that poor skill at (say) logical reasoning not only impairs one's performance at logical reasoning tasks but also impairs one's ability to evaluate one's own performance at logical reasoning tasks. You need to know that affirming the consequent is a logical error in order to realize that you've just committed a logical error in affirming the consequent. Otherwise, you're likely to think, "P implies Q, and Q is true, so P must be true. Right! Hey, I'm doing great!"

Although popular presentations of the Kruger-Dunning effect tend to generalize it to all skill domains, it seems unlikely that it does generalize universally. In domains where evaluating one's success doesn't depend on the skill in question, and instead depends on simpler forms of observation and feedback, one might expect more realistic self-evaluations by novices. (I haven't noticed a clear, systematic discussion of cases where Dunning-Kruger doesn't apply, though Kahneman & Klein 2009 is related; tips welcome.) For example: footraces. I'd wager that people who are slow runners don't tend to think that they are above average in running speed. They might not have perfect expectations; they might show some self-serving optimistic bias (Taylor & Brown 1988), but we probably won't see the almost flat line characteristic of Kruger-Dunning. You don't have to be a fast runner to evaluate your running speed. You just need to notice that others tend to run faster than you. It's not like logic where skill at the task and skill at self-evaluation are closely related.

So... what about ethics? Ought we to expect a moral Dunning-Kruger Effect?

My guess is: yes. Evaluating one's own ethical or unethical behavior is a skill that itself depends on one's ethical abilities. The least ethical people are typically also the least capable of recognizing what counts as an ethical violation and how serious the violation is -- especially, perhaps, when thinking about their own behavior. I don't want to over-commit on this point. Certainly there are exceptions. But as a general trend, this strikes me as plausible.

Consider sexism. The most sexist people tend to be the people least capable of understanding what constitutes sexist behavior and what makes sexist behavior unethical. They will tend either to regard themselves as not sexist or to regard themselves only as "sexist" in a non-pejorative sense. ("Yeah, so what, I'm a 'sexist'. I think men and women are different. If you don't, you're a fool.") Similarly, the most habitual liars might not see anything bad in lying or just assume that everyone else who isn't just a clueless sucker also lies when convenient.

It probably doesn't make sense to think that overall morality can be accurately captured in a single unidimensional scale -- just like it probably doesn't make sense to think that there's one correct unidimensional scale for skill at baseball or for skill as a philosopher or for being a good parent. And yet, clearly some baseball players, philosophers, and parents are better than others. There are great, good, mediocre, and crummy versions of each. I think it's okay as a first approximation to think that there are more and less ethical people overall. And if so, we can at least imagine a rough scale.

With that important caveat, then, consider the following possible relationships between one's overall moral character and one's opinion about one's overall moral character:

Dunning-Kruger (more self-enhancement for lower moral character):

[Note: Sorry for the cruddy-looking images. They look fine in Excel. I should figure this out.]

Uniform self-enhancement (everyone tends to think they're a bit better than they are):

U-shaped curve (even more self-enhancement for the below average):

Inverse U (realistically low self-image for the worst, self-enhancement in the middle, and self-underestimation for the best):

I don't think we really know which of these models is closest to the truth.

Rationally Speaking: Weird Ideas and Opaque Minds

What a pleasure and an honor to have been invited back to Julia Galef's awesome podcast, Rationally Speaking!

If you don't know Rationally Speaking, check it out. The podcast weaves together ideas and guests from psychology, philosophy, economics, and related fields; and Julia has a real knack for the friendly but probing question.

In this episode, Julia and I discuss the value of truth, daringness, and wonder as motives for studying philosophy; the hazards of interpreting other thinkers too charitably; and our lack of self-knowledge about the stream of conscious experience.

Introspecting an Attitude by Introspecting Its Conscious Face

In some of my published work, I have argued that:

(1.) Attitudes, such as belief and desire, are best understood as clusters of dispositions. For example, to believe that there is beer in the fridge is nothing more or less than to be disposed (all else being equal or normal) to go to the fridge if one wants a beer, to feel surprised if one were to open the fridge and find no beer, to conclude that the fridge isn't empty if that question becomes relevant, etc, etc. (See my essays here and here.)

And

(2.) Only conscious experiences are introspectible. I characterize introspection as "the dedication of central cognitive resources, or attention, to the task of arriving at a judgment about one's current, or very recently past, conscious experience, using or attempting to use some capacities that are unique to the first-person case... with the aim or intention that one's judgment reflect some relatively direct sensitivity to the target state" (2012, p. 42-43).

Now it also seems correct that (3.) dispositions, or clusters of dispositions, are not the same as conscious experiences. One can be disposed to have a certain conscious experience (e.g., disposed to experience a feeling of surprise if one were to see no beer), but dispositions and their manifestations are not metaphysically identical. Oscar can be disposed to experience surprise if he were to see an empty fridge, even if he never actually sees an empty fridge and so never actually experiences surprise.

From these three claims it follows that we cannot introspect attitudes such as belief and desire.

But it seems we can introspect them! Right now, I'm craving a sip of coffee. It seems like I am currently experiencing that desire in a directly introspectible way. Or suppose I'm thinking aloud, in inner speech, "X would be such a horrible president!" It seems like I can introspectively detect that belief, in all its passionate intensity, as it is occurs in my mind right now.

I don't want to deny this, exactly. Instead, let me define relatively strict versus permissive conceptions of the targets of introspection.

To warm up, consider a visual analogy: seeing an orange. There the orange is, on the table. You see it. But do you really see the whole orange? Speaking strictly, it might be better to say that you see the orange rind, or the part of the orange rind that is facing you, rather than the whole orange. Arguably, you infer or assume that it's not just an empty rind, that it has a backside, that it has a juicy interior -- and usually that's a safe enough assumption. It's reasonable to just say that you see the orange. In a relatively permissive sense, you see the whole orange; in a relatively strict sense you see only the facing part of the orange rind.

Another example: From my office window I see the fire burning downtown. Of course, I only see the smoke. Even if I were to see the flames, in the strictest sense perhaps the visible light emitted from flames is only a contingent manifestation of the combustion process that truly constitutes a fire. (Consider invisible methanol fires.) More permissively, I see the fire when I see the smoke. More strictly, I need to see the flames or maybe even (impossibly?) the combustion process itself.

Now consider psychological cases: In a relatively permissive sense, you see Sandra's anger. In a stricter sense, you see her scowling face. In a relatively permissive sense, you hear the shyness and social awkwardness in Shivani's voice. In a stricter sense you hear only her words and prosody.

To be clear: I do not mean to imply that a stricter understanding of the targets of perception is more accurate or better than a more permissive understanding. (Indeed, excessive strictness can collapse into absurdity: "No, officer, I didn't see the stop sign. Really, all I saw were patterns of light streaming through my vitreous humour!")

As anger can manifest in a scowl and as fire can manifest in smoke and visible flames, so also can attitudes manifest in conscious experience. The desire for coffee can manifest in a conscious experience that I would describe as an urge to take a sip; my attitude about X's candidacy can manifest in a momentary experience of inner speech. In such cases, we can say that the attitudes present a conscious face. If the conscious experience is distinctive enough to serve as an excellent sign of the real presence of the relevant dispositional structure constituting that attitude, then we can say that the attitude is (occurrently) conscious.

It is important to my view that the conscious face of an attitude is not tantamount to the attitude itself, even if they normally co-occur. If you have the conscious experience but not the underlying suite of relevant dispositions, you do not actually have the attitude. (Let's bracket the question of whether such cases are realistically psychologically possible.) Similarly, a scowl is not anger, smoke is not a fire, a rind is not an orange.

Speaking relatively permissively, then, one can introspect an attitude by introspecting its conscious face, much as I can see a whole orange by seeing the facing part of its rind and I can see a fire by seeing its smoke. I rely upon the fact that the conscious experience wouldn't be there unless the whole dispositional structure were there. If that reliance is justified and the attitude is really there, distinctively manifesting in that conscious experience, then I have successfully introspected it. The exact metaphysical relationship between the strictly conceived target and the permissively conceived target is different among the various cases -- part-whole for the orange, cause-effect for the fire, and disposition-manifestation for the attitude -- but the general strategy is the same.

[image source]

Rik Peels' Defense of the Reliability of Introspection

In a paper forthcoming in Philosophical Studies (also here), Rik Peels defends, against a variety of scientifically inspired objections, the traditional philosophical view that introspection is a reliable source of knowledge. He focuses especially on arguments I developed in my 2011 book Perplexities of Consciousness.

Peels considers arguments for the unreliability of introspection based on: the sometimes large effect of reporting method (e.g., button press vs. verbal report) on the content of introspective judgments; cultural variation in whether dreams are seen as colored; the poor correlation between self-reported visual imagery and performance on cognitive tasks often thought to be facilitated by visual imagery (such as mental rotation or folding tasks); people's ignorance of their capacity for echolocation; and people's ignorance of the lack of detail and precision in the visual field. In each case, Peels presents the skeptic's argument in a couple of pages and then offers a couple of pages of objections.

For example, Peels offers three responses to my observation that self-reports of mental imagery tend not to correlate with behavioral performance on seemingly related cognitive tasks. First, he points out that since people don't have access to other people's imagery, terms like "vivid" might be interpreted quite differently between subjects -- and thus people might be accurate in their own idiolect even if not in an outwardly measurable way. Second, he suggests that the aspects of mental imagery being reported (such as vividness) might not be relevant to the cognitive tasks at hand (like mental rotation), in which case lack of correlation is to be expected. Third, he notes that some studies (a minority) do report positive correlations. While in my discussion of this topic, I have given reasons to be leery of psychological findings that fail to replicate -- such as experimenter bias, "file-drawer" effects, the existence of confounding or intervening variables, and participants' tendency to confirm perceived research hypotheses -- Peels argues that I have presented insufficient positive evidence that such factors are at work in these cases.

The imagery case is, I think, illustrative of the debate in two ways: First, a lot hinges on examination of the empirical details, such as the precise nature of the measures. And second, a lot hinges on judgments of plausibility about which reasonable people might differ. For example, looking at hundreds of studies of visual imagery self-report and cognitive performance, a literature with lots of null results and non-replications, does one think (as I do), "That looks pretty bad for the reality of an underlying effect" or does one think (as Peels seems to), "There probably is an effect in there somewhere that the null studies aren't effectively getting at". This sort of thing is a matter of judgment. (And no, I don't think the typical statistical meta-analysis will yield a straightforward answer that we should take at face value.)

One's sense of plausibility, in such matters, comes close to being something like one's general academic worldview -- about philosophy, about psychology, and about their interaction. Here, I find the introduction and conclusion of Peels' essay interesting.

In the introduction, Peels lists four reasons the question of the reliability of introspection is important:

(a.) Ordinary "common sense" tends to treat introspection as a reliable source of knowledge. To reject its reliability is to challenge common sense.

(b.) There's a philosophical tradition from Descartes to Chalmers of taking introspection (appropriately restricted) as infallible. To reject introspection's reliability means abandoning the infallibilist tradition.

(c.)There's a contrary "scientistic" tradition that emphasizes science as the only secure source of knowledge and regards introspection as non-scientific. To reject introspection's reliability plays into the hands of those who want to privilege the epistemic role of science.

(d.) There's a debate within science about how to treat introspective self-reports.

As these observations make clear, issues about introspection are inextricable from one's general sense of what philosophers and psychologists should be doing. I suspect that it is background differences between Peels and me on these sorts of questions that drive our different senses of plausibility in interpreting the empirical details.

I want to resist the infallibilist tradition in philosophy, and the philosophical tradition that emphasizes "common sense", and other philosophical and metaphilosophical positions that seek to insulate philosophical reasoning from science. This is fundamental to my worldview and my vision of the nature of philosophy. My views on introspection are of a piece with this general worldview -- supporting that worldview, I think, but also supported by it. Other considerations that support such a worldview are (I argue) the incoherence and cultural variability of "common sense" and philosophical fashion, and evolutionary, cultural, and psychological reasons to think that people would likely be much worse reasoners about philosophical issues than they are about mundane, practical affairs.

I would add to Peels' four considerations also a fifth consideration of a somewhat different sort: People usually want to be taken at their word when they say they're not angry, not racist, happy with their life choices. To express doubt is to deny people a certain sort of authority over the story of what is going on in their own minds. An introspection-friendly approach tends to cede people that sort of self-authority; an introspective-skeptical approach denies that self-authority. Again, there are big-picture worldview questions at stake, which both inform one's sense of how to interpret the psychological evidence and are in turn either supported or undercut by some of that same evidence.

Although Peels' essay focuses on the empirical details of a few cases, this larger context both informs and motivates all work on the epistemology of introspection. Are we basically right about ourselves, and is philosophy of mind safe from radical scientific critique? Or is self-knowledge fragile and the armchair a tempting cozy spot to doze away into ignorance?

The Intrinsic Value of Self-Knowledge

April 2, I'll be a critic at a Pacific APA author-meets-critics session on Quassim Cassam's book Self-Knowledge for Humans. (Come!) In the last chapter of the book, Cassam argues that self-knowledge is not intrinsically valuable. It's only, he says, derivatively or instrumentally valuable -- valuable to the extent it helps deliver something else, like happiness.

I disagree. Self-knowledge is intrinsically valuable! It's valuable even if it doesn't advance some other project, valuable even if it doesn't increase our happiness. Cassam defends his view by objecting to three possible arguments for the intrinsic value of self-knowledge. I'll spot him those objections. Here are three other ways to argue for the intrinsic value of self-knowledge.

1. The Argument from Addition and Subtraction.

Here's what I want you to do: Imaginatively subtract our self-knowledge from the world while keeping everything else as constant as possible, especially our happiness or subjective sense of well-being. Now ask yourself: Is something valuable missing?

Now imaginatively add lots of self-knowledge to the world while keeping everything else as constant as possible. Now ask: Has something valuable been gained?

Okay, I see two big problems with this method of philosophical discovery. Both problems are real, but they can be partly addressed.

Problem 1: The subtraction and addition are too vague to imagine. To do it right, you need to get into details, and the details are going to be tricky.

Reply 1: Fair enough! But still: We can give it a try and take our best guess where it's leading. Suppose I suddenly knew more about why I'm drawn to philosophy. Wouldn't that be good, independent of further consequences? Or subtract: I think of myself as a middling extravert. Suppose I lose this knowledge. Stipulate again: To the extent possible, no practical consequences. Wouldn't something valuable be lost?

Alternatively, consider an alien culture on the far side of the galaxy. What would I wish for it? Would I wish for a culture of happy beings with no self-knowledge? Or, if I imaginatively added substantial self-knowledge to this culture, would I be imagining a better state of affairs in the universe? I think the latter.

Contrast with a case where addition and subtraction leave us cold: seas of iron in the planet's core. Unless there are effects on the planetary inhabitants, I don't care. Add or subtract away, whatever.

Problem 2: What these exercises reveal is only that I regard self-knowledge as something that has intrinsic value. You might differ. You might think: happy aliens, no self-knowledge, great! They're not missing anything important. You might think that unless some practical purpose is served by knowing your personality, you might as well not know.

Reply 2: This is just the methodological problem that's at the root of all value inquiries. I can't rationally compel you to share my view, if you start far enough away in value space. I can just invite you to consider how your own values fit together, suggest that if you think about it, you'll find you already do share these values with me, more or less.

2. The Argument from Nearby Cases.

Suppose you agree that knowledge in general is intrinsically valuable. A world of unreflective bliss would lack something important that a world of bliss plus knowledge would possess. I want my alien world to be a world with inhabitants who know things, not just a bunch of ecstatic oysters.

Might self-knowledge be an exception to the general rule? Here's one reason to think not: Knowledge of the motivations and values and attitudes of your friends and family, specifically, is intrinsically good. Set this up with an Argument from Subtraction: Subtracting from the world people's psychological knowledge of people intimate to them would make the world a worse place. Now do the Nearby Cases step: You yourself are one of those people intimate to you! It would be weird if psychological knowledge of your friends were valuable but psychological knowledge of yourself were not.

Unless you're a hedonist -- and few people, when they really think about it, are -- you probably thinks that there's some intrinsic value in the rich flourishing of human intellectual and artistic capacities. It seems natural to suppose that self-knowledge would be an important part of that general flourishing.

3. The Argument from Identity.

Another way to argue that something has intrinsic value is to argue that it is in fact identical to something that we already agree has intrinsic value.

So what is self-knowledge? On my dispositional view (see here and here), to know some psychological fact about yourself is to possess a suite of dispositions or capacities with respect to your own psychology. An example:

What is it to know you're an extravert? It's in part the capacity to say, truly and sincerely, "I'm an extravert". It's in part the capacity to respond appropriately to party invitations, by accepting them in anticipation of the good time you'll have. It's in part to be unsurprised to find yourself smiling and laughing in the crowd. It's in part to be disposed to conclude that someone in the room is an extravert. Etc.

My thought is: Those kinds of dispositions or capacities are intrinsically valuable, central to living a rich, meaningful life. If we subtract them away, we impoverish ourselves. Human life wouldn't be the same without this kind of self-attunement or structured responsiveness to psychological facts about ourselves, even if we might experience as much pleasure. And self-knowledge is not just some further thing floating free of those dispositional patterns that could be subtracted without taking them away too. Self-knowledge isn't some independent representational entity contingently connected to those patterns; it is those patterns.

You might notice that this third argument creates some problems for the straightforward application of the Argument from Addition and Subtraction. Maybe in trying to imagine subtracting self-knowledge from the world while holding all else constant to the extent possible, you were imagining or trying to imagine holding constant all those dispositions I just mentioned, like the capacity to say yes appropriately to party invitations. If my view of knowledge is correct, you can't do that. What this shows is that the Argument from Addition and Subtraction isn't as straightforward as it might at first seem. It needs careful handling. But that doesn't mean it's a bad argument.

Conclusion.

I'd go so far as to say this: Self-knowledge, when we have it (which, I agree with Cassam, is less commonly than we tend to think), is one of the most intrinsically valuable things in human life. The world is a richer place because pieces of it can gaze knowledgeably upon themselves and the others around them.

Three Kinds of Transparency in Self-Knowledge (Or Maybe 1 3/4 Kinds); and How the World Looks If You're Disagreeable

Transparency is a fun concept, as applied in the recent philosophical literature on self-knowledge (e.g., here; see also these past posts). I want to extend the transparency metaphor to self-knowledge of personality traits, where philosophers haven't yet taken it (to my knowledge).

First, the established versions:

Transparency 1: Sensory experience. As G.E. Moore and Gilbert Harman have noticed, when you try to attend to your visual experience you (usually? inevitably?) end up attending to external objects instead (or at least in addition). Attend to your experience of the computer screen. In some sense it seems like your attention slips right through the experience itself, lodging upon the screen.

Transparency 2: Attitudes. As Gareth Evans notices, when someone asks you if you think there will be a third world war, you typically answer by thinking about the outside world -- about the likelihood of war -- rather than by turning your attention inward toward your own mental states. More contentiously, thinking about whether you want ice cream also seems to involve, mostly, thinking about the world -- about the advantages and disadvantages of eating ice cream.

In both types of case, you learn about yourself by attending to the outside world. One feature of this metaphor that I especially like is that you can learn about distortions in yourself by noticing features of the world that don't align with your general knowledge. Gazing through a window, if the trees are a weird color, you know the window is tinted; if the trees wiggle around as you shift head position, you know the window is warped. If the lamplights at night radiate spears, you've learned something about distortions in your vision. If your paper is The Best Thing Ever Published, you've learned something about your egocentric bias.

This brings me to:

Transparency 3: Personality. Since I think personality traits and attitudes are basically the same thing, I regard Transparency 3 as a natural extension of Transparency 2. (And since 2 and 1 are also related, maybe we have 1 3/4 kinds of transparency rather than three kinds.) To find out if you're the kind of person who loves children, think about children. To find out if you're an extravert, think about parties. Since your personality colors your view of the world, one way to learn about your personality is to look at a relevant part of the world and notice its color. (You can also consider your past behavior; or directly try a label on for size; or ask friends for their frank opinion of you. Transparency-style reflection on the world isn't the only, or even always a very good, route to self-knowledge.)

This approach might work especially well for the "Big Five" personality trait Agreeableness. "Agreeable" people are those who self-rate, or are rated by others, as being concerned for other people, interested in them, sympathetic, helpful. Several recent studies suggest that people who self-rate as agreeable tend also to be more likely to rate others as agreeable (sympathetic, helpful, etc.), and also to rate other people positively in other ways too, especially if the other person is not very well known. If you're a sweetheart, the world seems to be full mostly of sweethearts and interesting people; if you're a jerk, the world seems to be full of jerks and fools. What I'm proposing to call the transparency approach to self-knowledge of personality simply involves running this observation the other direction: From the fact that the world seems full of uninteresting and disagreeable people, infer that you are a disagreeable person; from the fact that the world seems full of warm, loving, interesting people, infer that you are a relatively sympathetic, concerned person.

Now it would be really interesting if it turned out that these sorts of perceiver effects were actually better predictors of underlying personality as constituted by patterns of real-world behavior than are the typical self-rating scales used in personality psychology -- but it's probably too much to hope that the world would align so neatly with my own theoretical biases. (Hm, what does that last judgment reveal about my personality?)

Moral Self-Knowledge by Looking at Others' Faces

Our own moral character is largely unknown to us. Lots of jerks think they're just swell. Lots of saints and sweethearts suffer from moral self-doubt. But a formulaic inversion of one's moral self-opinion doesn't work either: Moral pride and moral self-condemnation sometimes fit the facts quite well. I conjecture approximately a zero correlation between people's moral self-opinions and their actual moral character.

Moral self-knowledge is an unruly beast that cannot, I think, be trapped and held still for systematic examination, partly because moral self-examination is itself a moral act, tangled up with the very traits under consideration. The jerk will tend toward a biased self-examination; the sweetheart will be biased in a different way; and the conclusions one habitually reaches, on either side, can reinforce or undercut the very traits self-ascribed. "Oh, I'm such a sweetheart, so much nicer than everyone else around me" is not, in most circumstances, a thought that sits very comfortably on its bearer.

The usual method by which we consider our moral character -- trying an adjective on for size, as it were, and asking, "Is that me? Am I trustworthy? Am I kind? Am I gentle?" -- is, as suggested by both informal observation and psychological research, a method of little probative value. Maybe somewhat better is taking an icy look at objective data about yourself or asking for the frank opinion of someone whose judgment you trust -- but both of those approaches are also seriously flawed.

So here's another approach to add to the stock -- an approach that is also flawed, but which deserves attention because its potential power hasn't yet, I think, been widely enough recognized. Look at the faces of the people around you. Central to our moral character is how we tend to view others nearby. The jerk sees himself as surrounded by fools and losers. The sweetheart vividly appreciates the unique talents and virtues of whomever he's with. The avaricious person sees the people around her as a threat to her resources (time, money, but also possibly space in the subway, position in line, praise from her peers). The person obsessed with social position sees people who vary finely in their relative social standing. Or consider: What do you notice about others' physical appearance? This reveals something morally important about you -- something not directly under your control, a kind of psychological tell.

Or so I think it's reasonable to suppose. I'm open to counterevidence, e.g., by experience sampling beeper methods, combined with some plausible related measures of moral character. But psychological science isn't there yet.

Of course you can game it. You can sit around and work yourself into a blissed-out appreciation of all those wonderful people around you, congratulating yourself on your sage-like moral awesomeness. This is a misfire, especially if there's an implicit (or explicit) comparative dimension to your moral self-assessment as you do this. (If only everyone were as good at me at seeing how wonderful everyone is!) Or you can choose to recall situations, or choose to put yourself in situations, disproportionately suggestive of the type of moral character you'd like to see yourself as having.

But I don't think it's inevitable that we game the method. I find it interestingly revealing (and disappointing) to look at strangers in the store or acquaintances at a party, letting my relatively uncensored assessments of them float up as an indication not of anything about them but rather as an indication of something about me, that I view them that way.

You can also notice things post-hoc: You can catch yourself viewing people in the way characteristic of the jerk, or in the way characteristic of the avaricious person, or of the person focused on status or sexual attractiveness.

It needn't always be negative. Sometimes you can congratulate yourself on being the one person in line who seemed to treat the cashier as a person. Sometimes you can feel good about the fact that you find yourself feeling good about the people around you. But I emphasize the negative for two reasons. First, I suspect that non-depressed people -- perhaps, especially, relatively affluent Western men? -- tend to err toward having too high an opinion of their moral character. Second, there's probably something cognitively or morally unstable, as I've gestured at a couple times above, about using this technique primarily for moral self-congratulation.

[Jeanette McMullin King has reminded me of the poem "The Right Kind of People", which fits nicely with this post.]

A Negative, Pluralist Account of Introspection

What is introspection? Nothing! Or rather, almost everything.

A long philosophical tradition, going back at least to Locke, has held that there is a distinctive faculty by means of which we know our own minds -- or at least our currently ongoing stream of conscious experience, our sensory experience, our imagery, our emotional experience and inner speech. "Reflection" or "inner sense" or introspection is, in this common view, a single type of process, yielding highly reliable (maybe even infallibly certain) knowledge of our own minds.

Critics of this approach to introspection have tended to either:

(a.) radically deny the existence of the human capacity to discover a stream of inner experience (e.g., radical behaviorism);

(b.) attribute our supposedly excellent self-knowledge of experience to some distinctive process other than introspection (e.g., expressivist or transparency approaches, on which "I think that..." is just a dongle added to a judgment about the outside world, no inward attention or scanning required); or

(c.) be pluralistic in the sense that we have one introspective mechanism to scan our beliefs, another to scan our visual experiences, another to scan our emotional experiences....

But here's another possibility: Introspective judgments arise from a range of processes that is diverse both within-case (i.e., lots of different processes feeding any one judgment) and between-case (i.e., very different sets of processes contributing to the judgment on different occasions) and yet also allows that introspective judgments arise partly through a relatively direct sensitivity to the conscious experiences that they are judgments about.

Consider an analogy: You're at a science conference or a high school science fair, quickly trying to take in a poster. You have no dedicated faculty of poster-taking-in. Rather, you deploy a variety of cognitive resources: visually appreciating the charts, listening to the presenter's explanation, simultaneously reading pieces of the poster, charitably bringing general knowledge to bear, asking questions and listening to responses both for overt content and for emotional tone.... It needn't be the same set of resources every time (you needn't even use vision: sometimes you can just listen, if you're in the mood or visually impaired). Instead, you flexibly, opportunistically use a diverse range of resources, dedicated to the question of what are the main ideas of this poster, in a way that aims to be relatively directly sensitive to the actual content of the poster.

Introspection, in my view, is like that. If I want to know what my visual experience is right now, or my emotional experience, or my auditory imagery, I engage not one cognitive process that was selected or developed primarily for the purpose of acquiring self-knowledge; rather I engage a diversity of processes that were primarily selected or developed for other purposes. I look outward at the world and think about what, given that world, it would make sense for me to be experiencing right now; but also I am attuned to the possibility that I might not be experiencing that, ready to notice clues pointing a different direction. I change and shape my experience in the very act of thinking about it, often (but not always) in a way that improves the match between my experience and my judgment about it. I have memories (short- and long-term), associations, things that it seems more natural and less natural to say, views sometimes important to my self-image about what types of experience I tend to have, either in general or under certain conditions, emotional reactions that color or guide my response, spontaneous speech impulses that I can inhibit or disinhibit. Etc. And any combination of these processes, and others besides, can swirl together to precipitate a judgment about my ongoing stream of experience.

Now the functional set-up of the mind is such that some processes' outputs are contingent upon the outputs of other processes. Pieces of the mind stay in sync with what is going on in other pieces, keep a running bead on each other, with varying degrees of directness and accuracy. And so also introspective judgments will be causally linked to a wide variety of other cognitive processes, including normally both relatively short and relatively circuitous links from the processes that give rise to the conscious experiences that the introspective judgments are judgments about. But these kinds of contingencies imply no distinctive introspective self-scanning faculty; it's just how the mind must work, if it is to be a single coherent mind, and it happens preconsciously in systems no-one thinks of as introspective, e.g., in the early visual system, as well as farther downstream.

[For further exposition of this view, with detailed examples, see my essay Introspection, What?]

An Objection to Group Consciousness Suggested by David Chalmers

For a couple of years now, I have been arguing that if materialism is true the United States probably has a stream of conscious experience over and above the conscious experiences of its citizens and residents. As it happens, very few materialist philosophers have taken the possibility seriously enough to discuss it in writing, so part of my strategy in approaching the question has been to email various prominent materialist philosophers to get a sense of whether they thought the U.S. might literally be phenomenally conscious, and if not why not.

To my surprise, about half of my respondents said they did not rule out the possibility. Two of the more interesting objections came from Fred Dretske (my undergrad advisor, now deceased) and Dan Dennett. I detail their objections and my replies in the essay in draft linked above. Although I didn't target him because he is not a materialist, [update 3:33 pm: Dave points out that I actually did target him, though it wasn't in my main batch] David Chalmers also raised an objection about a year ago in a series of emails. The objection has been niggling at me ever since (Dave's objections often have that feature), and I now address it in my updated draft.

The objection is this: The United States might lack consciousness because the complex cognitive capacities of the United States (e.g., to war and spy on its neighbors, to consume and output goods, to monitor space for threatening asteroids, to assimilate new territories, to represent itself as being in a state of economic expansion, etc.) arise largely in virtue of the complex cognitive capacities of the people composing it and only to a small extent in virtue of the functional relationships between the people composing it. Chalmers has emphasized to me that he isn't committed to this view, but I find it worth considering nonetheless, and others have pressed similar concerns.

This objection is not the objection that no conscious being could have conscious subparts (which I discuss in Section 2 of the essay and also here); nor is it the objection that the United States is the wrong type of thing to have conscious states (which I address in Sections 1 and 4). Rather, it's that what's doing the cognitive-functional heavy lifting in guiding the behavior of the U.S. are processes within people rather than the group-level organization.

To see the pull of this objection, consider an extreme example -- a two-seater homunculus. A two-seater homuculus is a being who behaves outwardly like a single intelligent entity but who instead of having a brain has two small people inside who jointly control the being's behavior, communicating with each other through very fast linguistic exchange. Plausibly, such a being has two streams of conscious experience, one for each homunculus, but no additional group-level stream for the system as a whole (unless the conditions for group-level consciousness are weak indeed). Perhaps the United States is somewhat like a two-seater homunculus?

Chalmers's objection seems to depend on something like the following principle: The complex cognitive capacities of a conscious organism (or at least the capacities in virtue of which the organism is conscious) must arise largely in virtue of the functional relationships between the subsystems composing it rather than in virtue of the capacities of its subsystems. If such a principle is to defeat U.S. consciousness, it must be the case both that

(a.) the United States has no such complex capacities that arise largely in virtue of the functional relationships between people, and

(b.) no conscious organism could have the requisite sort of complex capacities largely in virtue of the capacities of its subsystems.

Contra (a): This claim is difficult to assess, but being a strong, empirical negative existential (the U.S. has not even one such capacity), it seems a risky bet unless we can find solid empirical grounds for it.

Contra (b): This claim is even bolder. Consider a rabbit's ability to swiftly visually detect a snake. This complex cognitive capacity, presumably an important contributor to rabbit visual consciousness, might exist largely in virtue of the functional organization of the rabbit's visual subsystems, with the results of that processing then communicated to the organism as a whole, precipitating further reactions. Indeed turning (b) almost on its head, some models of human consciousness treat subsystem-driven processing as the normal case: The bulk of our cognitive work is done by subsystems, who cooperate by feeding their results into a global workspace or who compete for fame or control. So grant (a) for sake of argument: The relevant cognitive work of the United States is done largely within individual subsystems (people or groups of people) who then communicate their results across the entity as a whole, competing for fame and control via complex patterns of looping feedback. At the very abstract level of description relevant to Chalmers's expressed (but let me re-emphasize, not definitively endorsed) objection, such an organization might not be so different from the actual organization of the human mind. And it is of course much bolder to commit to the further view implied by (b), that no conscious system could possibly be organized in such a subsystem-driven way. It's hard to see what would justify such a claim.

The two-seater homunculus is strikingly different from a rabbit or human system (or even a Betelguesian beehead) because the communication is only between two sub-entities, at a low information rate; but the U.S. is composed of about 300,000,000 sub-entities whose informational exchange is massive, so the case is not similar enough to justify transferring intuitions from the one to the other.

Introspecting My Visual Experience "as of" Seeing a Hat?

In "The Unreliability of Naive Introspection" (here and here), I argue, contra a philosophical tradition going back at least to Descartes, that we have much better knowledge of middle-sized objects in the world around us than we do of our stream of sensory experience while perceiving those objects.

As I write near the end of that paper:

The tomato is stable. My visual experience as I look at the tomato shifts with each saccade, each blink, each observation of a blemish, each alteration of attention, with the adaptation of my eyes to lighting and color. My thoughts, my images, my itches, my pains – all bound away as I think about them, or remain only as interrupted, theatrical versions of themselves. Nor can I hold them still even as artificial specimens – as I reflect on one aspect of the experience, it alters and grows, or it crumbles. The unattended aspects undergo their own changes too. If outward things were so evasive, they’d also mystify and mislead.

Last Saturday, I defended this view for three hours before commentator Carlotta Pavese and a number of other New York philosophers (including Ned Block, Paul Boghossian, David Chalmers, Paul Horwich, Chris Peacocke, Jim Pryor).

One question -- raised first, I think, by Paul B. then later by Jim -- was this: Don't I know that I'm having a visual experience as of seeing a hat at least as well as I know that there is in fact a real hat in front of me? I could be wrong about the hat without being wrong about the visual experience as of seeing a hat, but to be wrong about having a visual experience as of seeing a hat, well, maybe it's not impossible but at least it's a weird, unusual case.

I was a bit rustier in answering this question than I would have been in 2009 -- partly, I suspect, because I never articulated in writing my standard response to that concern. So let me do so now.

First, we need to know what kind of mental state this is about which I supposedly have excellent knowledge. Here's one possibility: To have "a visual experience as of seeing a hat" is to have a visual experience of the type that is normally caused by seeing hats. In other words, when I judge that I'm having this experience, I'm making a causal generalization about the normal origins of experiences of the present type. But it seems doubtful that I know better what types of visual experiences normally arise in the course of seeing hats than I know that there is a hat in front of me. In any case, such causal generalizations are not sort of thing defenders of introspection usually have in mind.

Here's another interpretative possibility: In judging that I am having a visual experience as of seeing a hat, I am reporting an inclination to reach a certain judgment. I am reporting an inclination to judge that there is a hat in front of me, and I am reporting that that inclination is somehow caused by or grounded in my current visual experience. On this reading of the claim, what I am accurate about is that I have a certain attitude -- an inclination to judge. But attitudes are not conscious experiences. Inclinations to judge are one thing; visual experiences another. I might be very accurate in my judgment that I am inclined to reach a certain judgment about the world (and on such-and-such grounds), but that's not knowledge of my stream of sensory experience.

(In a couple of other essays, I discuss self-knowledge of attitudes. I argue that our self-knowledge of our judgments is pretty good when the matter is of little importance to our self-conception and when the tendency to verbally espouse the content of the judgment is central to the dispositional syndrome constitutive of reaching that judgment. Excellent knowledge of such partially self-fulfilling attitudes is quite a different matter from excellent knowledge of the stream of experience.)

So how about this interpretative possibility? To say I know that I am having a visual experience as of seeing a hat is to say that I am having a visual experience with such-and-such specific phenomenal features, e.g., this-shade-here, this-shape-here, this-piece-of-representational-content-there, and maybe this-holistic-character. If we're careful to read such judgments purely as judgments about features of my current stream of visual experience, I see no reason to think we would be highly trustworthy in them. Such structural features of the stream of experience are exactly the kinds of things about which I've argued we are apt to err: what it's like to see a tilted coin at an oblique angle, how fast color and shape experience get hazy toward the periphery, how stable or shifty the phenomenology of shape and color is, how richly penetrated visual experience is with cognitive content. These are topics of confusion and dispute in philosophy and consciousness studies, not matters we introspect with near infallibility.

Part of the issue here, I think, is that certain mental states have both a phenomenal face and a functional face. When I judge that I see something or that I'm hungry or that I want something, I am typically reaching a judgment that is in part about my stream of conscious experience and in part about my physiology, dispositions, and causal position in the world. If we think carefully about even medium-sized features of the phenomenological face of such hybrid mental states -- about what, exactly, it's like to experience hunger (how far does it spread in subjective bodily space, how much is it like a twisting or pressure or pain or...?) or about what, exactly, it's like to see a hat (how stable is that experience, how rich with detail, how do I experience the hat's non-canonical perspective...?), we quickly reach the limits of introspective reliability. My judgments about even medium-sized features of my visual experience are dubious. But I can easily answer a whole range of questions about comparably medium-sized features of the hat itself (its braiding, where the stitches are, its size and stability and solidity).

Update, November 25 [revised 5:24 pm]:

Paul Boghossian writes:

I haven't had a chance to think carefully about what you say, but I wanted to clarify the point I was making, which wasn't quite what you say on the blog, that it would be a weird, unusual case in which one misdescribes one's own perceptual states.

I was imagining that one was given the task of carefully describing the surface of a table and giving a very attentive description full of detail of the whorls here and the color there. One then discovers that all along one has just been a brain in a vat being fed experiences. At that point, it would be very natural to conclude that one had been merely describing the visual images that one had enjoyed as opposed to any table. Since one can so easily retreat from saying that one had been describing a table to saying that one had been describing one's mental image of a table, it's hard to see how one could be much better at the former than at the latter.

Roger White then made the same point without using the brain in a vat scenario.

I do feel some sympathy for the thought that you get something right in such a case -- but what exactly you get right, and how dependably... well, that's the tricky issue!

An Argument That the Ideal Jerk Must Remain Ignorant of His Jerkitude

As you might know, I'm working on a theory of jerks. Here's the central idea a nutshell:

The jerk is someone who culpably fails to respect the perspectives of other people around him, treating them as tools to be manipulated or idiots to be dealt with, rather than as moral and epistemic peers.

The characteristic phenomenology of the jerk is "I'm important and I'm surrounded by idiots!" To the jerk, it's a felt injustice that he must wait in the post-office line like anyone else. To the jerk, the flight attendant asking him to hang up his phone is a fool or a nobody unjustifiably interfering with his business. Students and employees are lazy complainers. Low-level staff failed to achieve meaningful careers through their own incompetence. (If the jerk himself is in a low-level position, it's either a rung on the way up or the result of injustices against him.)

My thought today is: It is partly constitutive of being a jerk that the jerk lacks moral self-knowledge of his jerkitude. Part of what it is to fail to respect the perspectives of others around you is to fail to see your dismissive attitude toward them as morally inappropriate. The person who disregards the moral and intellectual perspectives of others, if he also acutely feels the wrongness of doing so -- well, by that very token, he exhibits some non-trivial degree of respect for the perspectives of others. He is not the picture-perfect jerk.

It is possible for the picture-perfect jerk to acknowledge, in a superficial way, that he is a jerk. "So what, yeah, I'm a jerk," he might say. As long as this label carries no real sting of self-disapprobation, the jerk's moral self-ignorance remains. Maybe he thinks the world is a world of jerks and suckers and he is only claiming his own. Or maybe he superficially accepts the label "jerk", without accepting the full moral loading upon it, as a useful strategy for silencing criticism. It is exactly contrary to the nature of the jerk to sympathetically imagine moral criticism for his jerkitude, feeling shame as a result.

Not all moral vices are like this. The coward might be loathe to confront her cowardice and might be motivated to self-flattering rationalization, but it is not intrinsic to cowardice that one fails fully to appreciate one’s cowardice. Similarly for intemperance, cruelty, greed, dishonesty. One can be painfully ashamed of one’s dishonesty and resolve to be more honest in the future; and this resolution might or might not affect how honest one in fact is. Resolving does not make it so. But the moment one painfully realizes one’s jerkitude, one already, in that very moment and for that very reason, deviates from the profile of the ideal jerk.

There's an interesting instability here: Genuinely worrying about its being so helps to make it not so; but then if you take comfort in that fact and cease worrying, you have undermined the basis of that comfort.

Three Dimensions of Disagreement about Emotional Experience

William James's view of emotion is famous. Its most famous feature is his claim that emotional experience is entirely bodily:

If we fancy some strong emotion, and then try to abstract from our consciousness of it all the feelings of its characteristic bodily symptoms, we find we have nothing left behind... and that a cold and neutral state of intellectual perception is all that remains.... What kind of an emotion of fear would be left, if the feelings neither of quickened heart-beats nor of shallow breathing, neither of trembling lips nor of weakened limbs, neither of goose-flesh nor of visceral stirrings, were present, it is quite impossible to think. Can one fancy the state of rage and picture no ebullition of it in the chest, no flushing of the face, no dilatation of the nostrils, no clenching of the teeth, no impulse to vigorous action, but in their stead limp muscles, calm breathing, and a placid face? The present writer, for one, certainly cannot. The rage is as completely evaporated as the sensation of its so-called manifestations, and the only thing that can possibly be supposed to take its place is some cold-blooded and dispassionate judicial sentence, confined entirely to the intellectual realm, to the effect that a certain person or persons merit chastisement for their sins (1890/1950, vol. 2, p. 451-2).

Two other features are less commonly noted. One is that emotional experience is ever-present in rich detail:

Every one of the bodily changes, whosoever it be, is felt, acutely or obscurely, the moment it occurs. If the reader has never paid attention to this matter, he will be both interested and astonished to learn how many different local bodily feelings he can detect in himself as characteristic of his various emotional moods.... Our whole cubic capacity is sensibly alive; and each morsel of it contributes its pulsations of feeling, dim or sharp, pleasant, painful, or dubious, to that sense of personality that every one of us unfailingly carries with him (p. 451).

Another is that emotional experience is highly variable:

We should, moreover, find that our descriptions had no absolute truth; that they only applied to the average man; that every one of us, almost, has some personal idiosyncrasy of expression, laughing or sobbing differently from his neighbor, or reddening or growing pale where others do not.... The internal shadings of emotional feeling, moreover, merge endlessly into each other. Language has discriminated some of them, as hatred, antipathy, animosity, dislike, aversion, malice, spite, vengefulness, abhorrence, etc., etc.; but in the dictionaries of synonyms we find these feelings distinguished more by their severally appropriate objective stimuli than by their conscious or subjective tone (p. 447-448).

Disagreement continues, about all three issues.

Some scholars, such as Walter Cannon and Peter Goldie have argued that bodily sensations cannot possibly exhaust emotional experience; but others, such as Antonio Damasio and Jesse Prinz, have defended accounts of emotion that are broadly Jamesian in this respect.

Some scholars, such as John Searle (p. 140) have argued that we have ever-present emotional mood experiences even if they are often fairly neutral, while others, such as Russell Hurlburt and Chris Heavey, have argued that such feeling experiences are only present about 25% of the time on average. (This issue is a dimension of the larger question of how sparse or abundant human conscious experience is in general -- a question I have argued is methodologically fraught.)

I have seen less explicit discussion of how much variability there is in emotional experience between people, but some theories seem to imply that similar emotions will tend to have similar experiential cores: Keith Oatley and P.N. Johnson-Laird, for example, seem to think that each type of emotion -- e.g., anxiety, anger, disgust -- has a "distinctive phenomenological tone" (p. 34); and Goldie, while in some places emphasizing the complex variability of emotion, in other places (as in the article linked above), seems to imply that there's a distinctive qualitative character that an emotion like fear has which one cannot know unless one has experienced that emotion type. Hurlburt, in contrast, holds that people's emotional experiences are highly variable with no common core among them (e.g., here, p. 187, Box 8.8).

For all the work on emotion that has been done in the past 120 years, we are still pretty far, I think, from reaching a well-justified consensus opinion on these questions. Such is the enduring infancy of consciousness studies.

The Individual Differences Reply to Introspective Pessimism

I'm an introspective pessimist: I think people's introspective judgments about their stream of conscious experience are often radically mistaken. This is the topic of my 2011 book, Perplexities of Consciousness. Over the past few years, I've found the most common objection to my work to be what I'll call the Individual Differences Reply.

Many of my arguments depend on what I call the Argument from Variability: Some people report experiences of Type X, others report experience of Type not-X, but it's not plausible that Group 1 and Group 2 differ enough in their underlying stream of experience to make both claims true. Therefore, someone must be mistaken about their experience. So for example: People used to say they dreamed in black and white; now people don't say that; so someone is wrong about their dream experiences. Some people say that tilted coins in some sense look elliptical as though projected upon a 2D visual screen, while others say that visual experience is robustly 3D with no projective distortions at all. Some people say that objects seen 20 degrees from the point of fixation look fairly clear and colorful, while others say objects 20 degrees from center look extremely sketchy and indeterminate of color. Assuming a common underlying experience, disagreement reveals someone to be mistaken. (Fortunately, we needn't settle which is the mistaken party.)

The Individual Differences Reply challenges this assumption of underlying commonality. If people differ in their experience as radically as is suggested by their introspective reports, then no one need be mistaken! Maybe people did really used to dream in black and white and now they dream in color (see Ned Block's comment on a recent Splintered Mind post). Maybe some people really do see tilted coins as elliptical while others do not.

There are two versions of the Individual Difference Reply, which I will call the Stable Differences and the Temporary Differences versions. On the Stable Differences version, people durably differ in their experiences in such a way as to render their reports accurate as generalizations about their experiences. On the Temporary Differences version, people might be similar in their experiences generally, but when faced with the introspective task itself their experience shifts in such a way as to match their reports. For example, maybe everyone generally has similar experiences 20 degrees into the visual periphery, but when asked to introspect their visual experience, some people experience (and accurately report) clarity while others experience (and accurately report) haziness.

People who have pressed versions of the Individual Differences Reply on me include Alsmith forthcoming, Hohwy 2011, Jacovides 2012 and Hurlburt in Hurlburt & Schwitzgebel 2007.)

Here's how I think the dispute should be resolved: Look, on a case-by-case basis, for measurable brain or behavioral differences that tightly correlate with the differences in introspective report and which are plausibly diagnostic of the underlying variability. If those correlations are found, accept that the reports reveal real differences. If not, conclude that at least one party is wrong about their experience.

Consider black and white dreaming. Theoretically, we could hook up brain imagining machinery to people who report black and white dreams and to people who report color dreams. If the latter group shows substantially more neural activity in color-associated neural areas while dreaming, the reports are substantiated. If not, the reports are undermined. Alternatively, examine color term usage in dream narratives: How often do people use terms like "red", "green", etc., in dream diaries? If the rates are different, that supports the existence of difference; if not, that supports the hypothesis of error. In fact, I looked at exactly this in Chapter 1 of Perplexities, and found rates of color-term usage in dream reports to be the same during the peak period of black-and-white dream reporting (USA circa 1950) and recently (USA circa 2000).

Generally, I think the evidence often shows a poor relationship between differences in introspective report and differences in plausibly corresponding behavior or physiology.

Sometimes there is no systematic evidence, but antecedent plausibility considerations suggest against at least the Stable Differences version of the Individual Differences Reply: It seems highly unlikely that the people who report sharp visual experiences 20 degrees into the periphery differ vastly in general visual capacity or visual physiology from those who report highly indeterminate shape and color 20 degrees into the periphery. But that's an empirical question, so test it if you like!

What Are You Noticing When You Adjust Your Binoculars?

Maja Spener has written an interesting critique of my 2011 book, Perplexities of Consciousness, for a forthcoming book symposium at Philosophical Studies. My book is an extended argument that people have very poor knowledge of their own conscious experience, even of what one might have thought would be fairly obvious-seeming features of their currently ongoing experience (such as whether it is sparse or full of abundant detail, whether it is visually doubled, whether they have pervasive auditory experience of the silent surfaces around them). Maja argues that we have good introspective knowledge in at least one class of cases: cases in which our skillful negotiation with the world depends on what she calls "introspection-reliant abilities".

Maja provides examples of two such introspection-reliant abilities: choosing the right amount of food to order at a restaurant (which relies on knowledge of how hungry you feel) and adjusting binoculars (which relies on knowing how blurry the target objects look).

Let's consider, then, Maja's binoculars. What, exactly, are you noticing when you adjust your binoculars?

Consider other cases of optical reflection and refraction. I see an oar half-submerged in water. In some sense, maybe, it "looks bent". Is this a judgment about my visual experience of the oar or instead a judgment about the optical structure of my environment? -- or a bit of both? How about when I view myself in the bathroom mirror? Normally during my morning routine, I seem to be reaching judgments primarily about my body and secondarily about the mirror, and hardly at all about my visual experience. (At least, not unless we accept some kind of introspective foundationalism on which all judgments about the outside world are grounded on prior judgments about experience.) Or suppose I'm just out of the shower and the mirror is foggy; in this case my primary judgment seems to concern neither my body nor my visual experience but rather the state of the mirror. Similarly, perhaps, if the mirror is warped -- or if I am looking through a warped window, or a fisheye lens, or using a maladjusted review mirror at night, or looking in a carnival mirror.

If I can reach such judgments directly without antecedent judgments about my visual experience, perhaps analogously in the binoculars case? Or is there maybe instead some interaction among phenomenal judgments about experience and optical or objectual judgments, or some indeterminacy about what sort of judgment it really is? We can create, I'm inclined to think, a garden path between the normal bathroom mirror case (seemingly not an introspective judgment about visual experience) and the blurry binoculars case (seemingly an introspective judgment about visual experience), going either direction, and thus into self-contradiction. (For related cases feeding related worries see my essay in draft The Problem of Known Illusion and the Resemblance of Experience to Reality.)

Another avenue of resistance to the binoculars case might be this. Suppose I'm looking at a picture on a rather dodgy computer monitor and it looks too blue, so I blame the monitor and change the settings. Arguably, I could have done this without introspection: I reach a normal, non-introspective visual judgment that the picture of the baby seal is tinted blue. But I know that baby seals are white. So I blame the monitor and adjust. Or maybe I have a real baby seal in a room with dubious, theatrical lighting. I reach a normal visual assessment of the seal as tinted blue, so I know the lighting much be off and I ask the lighting techs to adjust. Similarly perhaps, then, if I gaze at a real baby seal through blurry binoculars: I reach a normal, non-introspective judgment of the baby seal as blurry-edged. But I know that baby seals have sharp edges. So I blame the binoculars and adjust. Need this be introspective at all, need it concern visual experience at all?

In the same way, maybe, I see spears of light spiking out of streetlamps at night -- an illusion, some imperfection of my eyes or eyewear. When I know I am being betrayed by optics, am I necessarily introspecting, or might I just be correcting my normal non-introspective visual assessments?

This is a nest of issues I am struggling with, making less progress than I would like. Maybe Maja is right, then. But if so, it will take further work to show.

Two Types of Hallucination

Oliver Sacks is one of the great essayists of our time. I have just finished his book Hallucinations.

Sacks does not, I think, adequately distinguish two types of hallucination. I will call them doxastic and phenomenal. In a phenomenal hallucination of A, one has sensory experience as of A. In a doxastic hallucination, one thinks one has sensory experience as of A. The two can come apart.

Consider this description, from page 99 of Hallucinations (and attributed to Daniel Breslaw via David Ebin's book The Drug Experience).

The heavens above me, a night sky spangled with eyes of flame, dissolve into the most overpowering array of colors I have ever seen or imagined; many of the colors are entirely new -- areas of the spectrum which I seem to have hitherto overlooked. The colors do not stand still, but move and flow in every direction; my field of vision is a mosaic of unbelievable complexity. To reproduce an instant of it would involve years of labor, that is, if one were able to reproduce colors of equivalent brilliance and intensity.

Here are two ways in which you might come to believe the above about your experience: (1.) You might actually have visual experiences of the sort described, including of colors entirely new and previously unimagined and of a complexity that would require years of labor to describe. Or (2.) you might shortcut all that and simply arrive straightaway at the belief that you are undergoing or have undergone such an experience -- perhaps with the aid of some unusual visual experiences, but not really of the novelty and complexity described. If the former, you have phenomenally hallucinated wholly novel colors. If the latter, you have merely doxastically hallucinated them.

The difference seems important -- crucial even, if we are going to understand the boundaries of experience as revealed by hallucination. And yet phenomenal vs. merely doxastic hallucinations might be hard to distinguish on the basis of verbal report alone, and almost by definition subjects will be apt to confuse the two. I can recall no point in the book where Sacks displays sensitivity to this issue.

Once I was attuned to it, the issue nagged at me again and again in reading:

Time was immensely distended. The elevator descended, "passing a floor every hundred years" (p. 100).

Then my whole life flashed in my mind from birth to the present, with every detail that ever happened, every feeling and thought, visual and emotional was there in an instant (p. 102).

I have had musical hallucinations (when taking chloral hydrate as a sleeping aid) which were continuations of dream music into the waking state -- once with a Mozart quintet. My normal musical memory and imagery is not that strong -- I am quite incapable of hearing every instrument in a quintet, let alone an orchestra -- so the experience of hearing the Mozart, hearing every instrument, was a startling (and beautiful) one (p. 213).

The possibility of merely doxastic hallucination might arise especially acutely when subjects report highly unusual, almost inconceivable, experiences or incredible detail beyond normal perception and imagery; but of course the possibility is present in more mundane hallucination reports too.

(A fan of Dennett might suggest that there's no difference between the phenomenal and doxastic hallucinations; but I don't know what Dennett himself would say -- probably something more complex than that.)