Top-down and Bottom-up Concepts

A response to "Is moral knowledge possible?"

Photo by Glen Carrie on Unsplash

Lorenzo Elijah has written a blog post purporting to show how moral knowledge is possible. This is a response to it, so go read it first here.

The argument draws on Goff’s distinction between transparent and opaque concepts. According to Goff, phenomenal concepts are transparent. Transparent concepts are ones that allegedly “Reveal the essence of the referent to anyone who grasps them.” Triangles serve as an example. I already don’t grant this, as I think it gets the nature of such concepts wrong.

I think the concept of a triangle is a stipulated, rule-bound notion that we’ve constructed, whereby the content of the concept is one that emerged out of a particular discipline with a fixed, intentionally constructed foundation. Compare this to the rules of a game. It is “phenomenally transparent” that bishops move diagonally and rooks move in a straight line in chess. As soon as I understand the concept of bishops and rooks, I “fully grasp their essence.” Only, all this amounts to is simply learning that what it means to be a bishop or rook just is to exhibit these traits. Likewise for triangles.

Conversely, “opaque concepts” allegedly don’t reveal their essence to anyone who grasps them. Again, I don’t think this is a genuine dichotomy; I don’t think there are concepts that (for some mysterious reason) are a bit voyeuristic and strip down so we can see all their tidbits, while others are more coy and waltz about clothed, revealing only an ankle here or a sultry gaze over there. Rather, I think Goff and others are conflating this distinction with a different distinction. Some of the language we use refers to “top-down” concepts, while other language refers to “bottom-up” concepts.

When we are personally in charge of the content of a concept, it has a top-down structure where we start with the concept, then slap a label onto it. They aren’t “transparent.” They’re stipulated. This puts us fully in charge of the contents. When this occurs, we go from the stipulated content of the concept to the terms we use to refer to the content of the concept.

Content of the concept → Label or name for the concept

In short, top-down concepts are defined by and characterized fully in terms of stipulative accounts of their meaning. These concepts typically emerge from technical disciplines in which we take some set of characteristics and pin a name on some useful notion with clear and distinct boundaries, e.g., “we need a name for the geometric figure that plays this role in Euclidean geometry so let’s call it a ‘triangle’.” Obviously I don’t think this is exactly how things happened; there’s more of a dynamic feedback between the needs of some technical field and the terms it uses, and terms gradually become entrenched over some span of time, but roughly the idea is that we pin labels to concepts that we ourselves conceive of as having a fixed set of characteristics.

Bottom-up concepts reverse this structure. We don’t stipulate their contents. Rather, these “concepts” pick out things we experience. Long before anyone sat down to work out what water was, we were drinking water, swimming in water, and so on. So we go from the stuff we refer to to the content:

The thing we refer to with a given name/label → the content of the concept

We don’t stipulate what the content of “water” is because it’s not up to us. It’s that stuff out there. It is whatever it turns out to be. The name “bottom-up” is a bit misleading, since the “concepts” in question arguably aren’t concepts or are best construed as a set of concepts comprised of a family of notions that overlap in meaning. Rather, these “concepts” are roughly picked out by shared linguistic, philosophical, and other overlapping practices centered on our shared sensory experiences. Bottom-up concepts emerge in ordinary language. Ordinary language is largely improvisational, with words shading and varying in a vast, semi-overlapping panoply of context-distinct uses. As such, there is no distinctive, rigid stipulative use for a given term or concept. People won’t agree on what counts as a weapon, or a cake, or a sandwich, not because the content of the concept isn’t transparent, but because our ordinary linguistic practices aren’t top-down and stipulative. They’re improvisational, dynamically evolving across contexts and over time, and needs-driven. We will bend and warp language to serve our purposes as needed. Where top-down concepts are a rigid rod of iron, bottom-up concepts are amorphous, clay-like, and ready to be molded to suit the many practical purposes of everyday life.

Analytic philosophy goes astray in large part due to its failure to grapple with the way language and human thought actually work. In its effort to regiment the way people speak in accord with the normative standards and theoretical presuppositions endemic to the field, it engenders a systematic failure to adopt the appropriate tools for examining ordinary language and the “concepts” that emerge from it. Instead, ordinary language is treated almost like a mysterious code: terms pick out “concepts,” and the concepts have “essences” or content that isn’t immediately apparent, but must be discovered via careful armchair reasoning.

I don’t think ordinary language is like this. I don’t think ordinary terms specifically refer to some particular “concept,” some of which have content we can “see,” while others have content we have to go and discover (such as by using the tools of science).

Note how Goff characterizes concepts as “revealing” their essences, while opaque concepts don’t. This framing of the distinction makes it seem as though there’s some abstract repository of concepts whose content we access and discover via different methods: the “essence” of some concepts is somehow “acquired” because the concept “reveals” itself to us (however that’s supposed to be cashed out epistemologically, psychologically, and so on); as a consequence, some concepts are promiscuous, while others play hard to get. Compare this instead to my distinctions, in light of the examples Elijah provides:

A triangle is a transparent concept. The concept of a triangle refers to any closed object composed of three straight lines. Simply by knowing this concept, you know the essence of a triangle. You can figure out that the internal angles of a triangle must sum to 180 degrees just by thinking.

Water is an opaque concept. Simply knowing the concept of water doesn’t mean you can figure out that water is essentially H2O. That requires scientific investigation.

The former, triangle, is a top-down concept the content of which was specified by us. The latter, water, isn’t. Water is an ordinary language term we use to refer to a recurring experience we have of particular substances we encounter. Its content isn’t invented by us, but must be acquired via scientific investigation. However, both concepts are ultimately rooted in our experiences and practices. The concept of a triangle would not have emerged in the absence of our direct engagement with and experience of the world. Navigating that world can be facilitated in part by certain constructed, useful idealizations, which is how geometry emerged. Math is a human invention, or tool, we ourselves created. Water isn’t. That’s why water’s “essence” is “opaque.” We use the term water to pick out a cluster of experiences. If we wanted to, we could invent an idealization, or top-down concept based on our acquired knowledge of chemistry, whereby we stipulate that water is H₂O. Or we could just not do this. It’s up to us. If we did, then the concept of “water” in this technical, stipulative concept, would be “transparent.” In fact, its content very likely is transparent in certain academic contexts, where “water” is just an easy way to say H₂O. Maybe this is how chemists talk when doing chemistry, and they don’t fret over or find the notion of what water “really” is to be remotely mysterious. Incidentally, the “water” we typically encounter isn’t pure H₂O since it has impurities (though one might argue that we’re not referring to these when we refer to water).

In any case, on Goff’s view, pain is a transparent concept. Now, Goff isn’t the one doing the talking here. Elijah is. And Elijah continues:

This distinction in hand, Goff argues our phenomenal concepts are transparent. Roughly, his argument begins with the Cartesian thought that I can be virtually certain of my direct, immediate experiences.

When I step on a Lego brick, I feel the sensation of blinding pain. Is it possible for me to doubt that in that very moment, I am feeling pain? Goff thinks I can’t. In his words, I’m super justified in believing there is pain going on. This belief is super justified because I know it with the same level of certainty that I know 2 + 2  =  4 (at least in the moment).

I don’t agree with this. I think you could doubt you’re in pain. I have a chronic pain condition. I am almost always in pain. It’s a dull, low-level pain that varies in intensity. Sometimes it’s distracting and unpleasant. Usually it’s just kind of there, slightly worsening my life. But sometimes it’s mostly (or perhaps even entirely) suppressed. I go long periods of time where my attention is directed elsewhere, and the pain is “tuned down” in my conscious awareness. Since I am so habituated to the presence of the pain, I often just presume it is there. But sometimes these background differences in intensity vary without my noticing, and when I redirect my attention to the pain, I have to really consider whether I’m “in pain” or not. I’ve had experiences ranging from the following:

1. “Yep, there it is. Ugh.”

2. “Oh, yea, there it is. But just barely. Nice.”

3. “Is that the pain? I think so. Yea. Yea it is.”

4. “Is the pain there? Where? Hmm…well, I can’t seem to find it. I guess it’s gone. Awesome!”

Given my lifelong familiarity with pain, if I were confident about anything, it’s that I am not always confident whether I am “in pain.” Some might even argue that the experiences I’ve described don’t even constitute pain. Pain, one might insist, must prompt an aversive reaction or must exhibit a distinct phenomenology of displeasure. Since there are instances in which my pain doesn’t do either of these, it’s “not really pain.”

Ha! Well, look what we have here. What we have, in such cases, is a philosopher’s attempt at stipulating a top-down definition. Pain, so they say, “just is” the kind of experience that prompts an aversive phenomenology. If what I’m experiencing doesn’t have this quality, it’s just not pain. Note how this is analogous to simply stipulating that water is H₂O, so whatever I’m drinking, if it’s not H₂O, it isn’t water.

This kind of move is the subtle shift, the sleight of hand, philosophers often employ: a shift between our ordinary use of terms like “pain,” and some stipulative, top-down, theoretical construction on the part of the philosopher. The way I personally use the term “pain,” it can and does encompass aspects of my mental life that I can doubt.

However, the example Elijah gives is stepping on a Lego brick, a notorious source of “blinding pain.“ Maybe “blinding” pain can’t be doubted. This still doesn’t convince me. I don’t know enough about the experience of blinding pain to know if it’s possible to doubt it or not, and I don’t know if this is an a priori question. But setting such meta-doubt aside (i.e., my doubt about whether it’s possible to doubt one is in blinding pain), another issue is that it does not follow from the fact that one is incapable of doubting they’re in pain that they are “justified,” much less “super justified” in believing they’re in pain. Incorrigibility is not a uniquely powerful form of justification. If it were, then the world’s most obstinate idiots, the kinds of people utterly incapable of changing their minds about anything, would be …what? Ultra justified merely in virtue of their intransigence? Yet we’re told:

This belief is super justified because I know it with the same level of certainty that I know 2 + 2  =  4 (at least in the moment).

This seems to conflate confidence with justification. 2+2=4 is true because we’ve defined terms in such a way that it’s true. One’s private experiences of pain aren’t like this. We all know how the pieces move in chess because we’ve mutually agreed that they do. This is a matter we can publicly check on with others. But there’s no way to check on the private experiences of Goff or anyone else to determine, from the outside, whether they even have the experiences that they report having, much less that they are of some kind that we (from the outside) should consider whatever judgments they arrive at on the basis of their alleged experiences to be “justified.” If Goff or anyone else’s personal confidence or incorrigibility “justifies” their beliefs, it is a hollow justification because it is entirely private, and utterly inaccessible to the rest of us.

Compare, for example, a prosecution in a criminal trial claiming that they had “private justification” of the defendant’s guilt, and thus couldn’t present it for the jury’s consideration. Nobody would take this seriously. Likewise, I see little reason to take anyone’s private justification seriously unless there is some external method of corroborating the reliability of the judgments in question.

Now, returning to the matter of pain, Elijah explains why we’re “super justified” in our judgments about pain:

Goff’s answer is that super justification is best explained by transparency. If the whole nature of pain is present in your experience (because pain just is that experience), then you grasp the nature of pain once you grasp the concept. So, phenomenal concepts are transparent to us.

Since I don’t grant that we’re super justified in the first place, I don’t grant that we need to explain why we are super justified. I’d first need to see a good argument for why we are “super justified.” In the meantime, I also reject the notion that “pain” is a transparent concept. I don’t grant that this is a real kind of concept in the first place. Going by my own distinctions, is the meaning of pain stipulated? In that case, if it includes content I don’t agree exist, e.g., that pain is “intrinsically bad,” then I don’t think this is something someone could experience and I deny anyone has “pain” of the relevant kind. If it’s instead a bottom-up concept, then its content isn’t stipulated and is instead something we’d have to figure out, so Goff would be mistaken.

Already, then, there are multiple points where I don’t think Goff’s argument gets off the ground. I don’t grant the transparent/opaque distinction, I don’t grant that we’re “super justified” about our judgments about pain, and I don’t grant that pain is a transparent concept. I don’t think we’ve been given good reasons, here at least, for endorsing any of these claims.

Nevertheless, Elijah continues:

Suppose we accept Goff’s argument for the transparency of phenomenal concepts. If so, then once I grasp the concept of pain, I know the essence of pain. So what?

Well, notice that pain is not neutral, pain is bad. Badness is a negatively valanced normative concept. To say that something is bad, is to say you have reason to avoid it. When I stub my toe, I immediately recognise the badness of my pain.

I see more moves being made here that I don’t grant. I don’t grant that to say that something is bad is to say “you have a reason to avoid it.” This is ambiguous. What does Elijah mean by this? I don’t know. If it’s a claim consistent with my own view of reasons, then I grant this, but it isn’t going to get you to moral realism. If it isn’t, and Elijah is instead invoking some other notion of reasons, then I would reject this claim, leading to yet another point of departure. Elijah also says

I immediately recognise the badness of my pain.

This phrasing worries me. It sounds like it is reifying the notion of something being bad, as though “badness” is some feature of the pain itself, out there in the world, distinct from our evaluations. But this isn’t how I view things being good or bad. I don’t think there literally is any such thing as badness, and I don’t think we “recognize” it as something apart from ourselves. Rather, I think goodness and badness are fundamentally matters of our personal appraisal, or terms we use to refer to the real or hypothetical appraisals of others. As such, the language of recognizing “the badness” of pain employs language that I’d keep an eye on, because it is worded in a way that hints at a conception of pain that I also don’t grant. Elijah continues:

What’s more, if I think about different scenarios with similar pain, I get the intuition that these experiences of pain would be bad too. Indeed, I can’t think of a single possible world with pain like this in which I don’t see its badness.

This is too ambiguous to evaluate. What kinds of scenarios is Elijah referring to? Scenarios where Elijah himself is experiencing pain, or anyone? Because it does not follow from the fact that we find our own pain to be bad that therefore anyone else’s pain is bad. If the intuition is that it’s “bad” for someone else to be in pain other than Elijah, not relative to Elijah’s standards, but in some stance-independent way, well…I don’t have that intuition. In other words, it doesn’t seem to me that it is stance-independently bad for anyone to be in pain, myself included. I also worry about the notion of “seeing” the badness of pain. I’m not quite sure how literally we should take such language. I don’t think anyone can “see” badness. Elijah continues:

Now what explains these intuitions? I think the best explanation is that badness is part of pain’s essence. In other words, badness is part of what it is to be in pain.

Since I don’t know what the content of these intuitions is, it’s hard to offer an explanation. Does Elijah think pain is stance-independently bad, or bad in some other way? If it’s the latter, then these intuitions won’t support an argument for moral realism. If it’s the former, it’s not clear to me whether such intuitions are justified, much less super justified. Either way, I don’t share these intuitions, and, as I’ve already highlighted, there are so many points in the case made here where I’d disagree, and that I don’t think we’ve been given much in the way of arguments or reasons to endorse such a view, that I see very little going for this account.

Elijah ends with a few other remarks. Here’s one:

This argument won’t convince anyone to become a moral realist. But so what? The argument isn’t even trying to do that. Its aim is to show how moral knowledge is possible, not to convince sceptics.

I don’t think the argument shows that moral knowledge is possible, either. At least not in a way that’s interesting or worth engaging with. Too much ambiguity. Too many claims made without arguments or evidence in their favor. It’s unclear whether we’ve been shown that moral knowledge is possible. And, in any case, I don’t think pain alone is an adequate foundation for a plausible moral theory.