[I’ve succumbed to the urge to use italics in this, although my original aim was to express these thoughts as a thread of Tweets. — Jeremy]
There is a variety of different kinds of ought. For example, rationally, “anyone who believes that p and that p implies q ought to believe that q”. Prudentially, “we ought to take vitamin D daily”. In chess, “players ought to get their opponent’s king into checkmate”. And so on.
Of course there are moral oughts as well. But focusing on moral oughts tends to cloud our judgement on the nature of oughts in general, because most of us yearn for something we can’t have: “moral facts”. To grasp the true nature of oughts, let’s accept that they are non-factual.
An ought enjoins a particular sort of behaviour. It might explicitly or implicitly contain a description of the sort of behaviour it enjoins, or explicitly or implicitly contain a description of a goal — the state of affairs the enjoined behaviour promotes.
In other words, oughts have “contents”. This means they have something in common with other claims that have contents, such as ascriptions of various “propositional attitudes”.
The basic propositional attitudes are belief and desire, whose contents correspond to declarative sentences. Schematically, Ab(p) — “agent A believes that p”, Ad(p) “agent A desires that p”. Other propositional attitudes (hope, fear, etc.) are combinations of belief and desire.
An ought takes the form a(p) — “act so that p”, addressed to no one in particular. It’s a content-bearing claim, but it’s not a propositional attitude. It doesn’t have truth conditions, and it stands or falls independently of whether or not any agents comply with it.
Although the content sentence p of an ought is true or false, and an attribution of belief or desire to particular agent A is true or false, an ought is neither true nor false. Oughts aren’t connected to specific agents — whom they are addressed to is a matter of context.
All oughts purport to have an obliging or binding force on those they are addressed to — they prescribe, exhort, demand or direct behaviour. This purportedly binding aspect distinguishes oughts from merely factual claims, which just describe what happens to be the case.
An ought has binding force only if it gives its addressee a reason to act. Such a reason is a combination of beliefs (about facts) and a desire (for a goal). If there is wide factual agreement between issuer and addressee, the crucial ingredient is a shared desire/goal.
Oughts are addressed to genuine agents — beings that act and have reasons to act. At the very least, they must have beliefs and desires and the ability to “choose” how to act, even if this amounts to nothing more than a stronger desire overruling weaker desires.
Some agents (such as thermostats) are so rudimentary, their goal-directed states hardly count as desires at all — or not full desires, anyway. Very well then, think of those states as “sub-desires” of “sub-agents”. It is a matter of degree, not of kind.
Rudimentary agents also have “sub-beliefs” — belief-like states which co-vary reliably with states of the world. (The co-variance is non-semantic information — semantic information enters the picture with interpretation and the attribution of propositional attitudes.)
Oughts extend to other agents somewhat as true beliefs extend to the facts that make them true. The analogy is not exact, but it’s close enough to rule out talk of “my ought” — it misses the point in much the same way as talk of “my truth”.
Addressees must be already committed to a shared goal for any ought to have leverage over them — in the above examples, the goals are of believing truths, of staying healthy, and of winning at chess. (To someone who wants to die young, “you ought not to smoke” has no leverage.)
If the issuer of an ought and its addressee already share a goal, they needn’t always strive harmoniously to achieve the same state of affairs — consider opponents in chess. But nothing further need be done for an ought expressing the shared goal to bind them both.
What truth is to belief, satisfaction is to desire. A desire is satisfied when its goal is realized, whether or not the agent knows that it is realized. So in the present context the word ‘satisfaction’ has nothing much to do with “a feeling of satisfaction”.
Below, I refer to the “broadness” of desires and “scope” of their goals, meaning the range of states of affairs that would satisfy them. (Analogously, the scope of the sentence ‘The cat is on the mat’ is narrower than the scope of the sentence ‘A cat is on the mat’.)
It is possible to implant a new desire (and thereby a new goal) in an agent, by using persuasive enough evidence to change their background beliefs, combined with an appeal to a broader desire that the agent has already, with a goal of wider “scope” than the original goal.
For example, suppose you want to go to Starbucks for coffee. I could make you want to go to Costa instead, if I could persuade you that Starbucks is closed. I’d also have to appeal to a broader desire that you have already — for any kind of coffee — with its goal of wider scope.
I could implant a further new desire, if I could persuade you that no coffee is available anywhere nearby, and could appeal to a goal of still wider scope, the object of a yet broader desire: for a hot drink… And so on, through ever-broader desires for goals of wider scope.
This pattern of appealing to goals of ever-wider scope mirrors that of persuasion by appealing to factual claims of increasing generality, eventually reaching laws — the most general sort of factual claim. Laws “confer warrant”, by assuring us that something “must” be the case.
For example, suppose I throw a ball up into the air. We know that “what goes up must come down” — it’s a law-like generality that admits of no exceptions. So this particular ball “must” come down again, sooner or later.
We needn’t take the “must” here to express any sort of “natural necessity”. We use the word ‘must’ because the law and initial conditions yield an expectation, which is in fact fulfilled — much as we say “that must be him!” if we’re expecting a visitor and the doorbell rings.
As well as being general, laws are systemically important to an agent’s belief system — lawlike beliefs are the ones we are most reluctant to give up in the face of new evidence. Likewise, the widest goals are systemically important to agents’ systems of desire.
We might give up the goal of having a cup of Starbucks coffee for the wider goal of having any sort of coffee, and that for the goal of having a hot drink, and that for the goal of being refreshed… The wider the goal we retreat to, the more unshakeable it becomes.
Oughts have a “flavour of necessity”, whose strength depends on the scope of the shared goals that underwrite them. This is different from a “flavour of universality”, whose strength depends instead on how widespread the goal is in the population at large.
The retreat to ever-broader desires reflects the shrub-like structure of any desire system. We start off at a young outer shoot and work our way down towards one of relatively few ground-level stems of our most basic values: love, truth, beauty, loyalty, morality, comfort, etc.
Extending this metaphor, beneath ground level basic stems are attached to a single root, life’s “prime directive”: promote the proliferation of your genes in future generations. This is the source of all basic desires, urges, values, goals, etc. (but it is not itself a desire).
The generality of a law confers a warrant to believe, and likewise, the width of scope of a goal confers — or at least purports to confer — the “to-be-done-ness” of a summons to act. The “ought” of the latter is analogous to the “must” of the former.
New beliefs can change intermediate desires aimed at goals of narrow scope, as we saw above, but they have less leverage over broader desires aimed at goals of wider scope. The wider the scope of a goal, the more resistant it is to “new information”.
The thought that oughts can be derived from ises alone is partly inspired by the idea that new information can change intermediate desires, as above. If facts can change desires, the idea seems to go, facts might be able to posit goals as well.
But oughts derive their to-be-done-ness from broader desires aimed at goals of wider scope, and these are resistant to such change. Beliefs and reasoning can’t change the most basic desires — the purpose of reason is to serve those desires by helping to realize them, as Hume saw.
The thought that oughts can be derived from ises alone is also inspired by the idea that having a desire is a fact about an agent. A factual inquiry might seem to reveal an agent’s “own oughts”, but these lack the binding power that characterizes genuine oughts.
It’s a fact about chess that the object of the game is to get the opponent’s king into checkmate, and it’s a fact about chess players that they are guided by that goal. The ought that expresses the goal does not describe any such facts, but instead prescribes behaviour.
The ought that expresses the object of the game of chess has binding power over someone who is committed to playing proper chess, but not over someone who wants to end a game quickly by playing “suicidal chess” (i.e. deliberately sacrificing his own men).
Such a player would be wholly compliant with all of the “factual” rules of chess — which describe how each piece can move, how they are arranged at the start of the game, etc. — but he is not playing proper chess, because he is not bound by the non-factual “ought of the game”.
Oughts are expressions of desire rather than descriptions of desire. Mere descriptions of an agent’s psychological states can’t yield genuine oughts, because their distinctive feature is binding force, and that can only apply to agents who share the relevant desire.
Desires are real psychological states of agents, but their goals (states of affairs that would satisfy them) usually don’t exist — desires point to a lack, need or want. By contrast, inasmuch as oughts actually direct behaviour, they often point towards real states of affairs.
For example, the rules of the road say that drivers ought to stop for pedestrians at zebra crossings. And they usually do. So the ought points towards innumerable facts — the countless occurrences of actual drivers routinely stopping for pedestrians.
Because many familiar oughts are usually or “normally” followed like that, the word ‘norm’ is often used as a synonym for ‘ought’ (used as a noun). This is misleading, because the word ‘ought’ is prompted here by a mere expectation that agents will act to bring it about that p.
That is an epistemic expectation on the part of the observer, underwritten by enumerative induction, rather than an ought expressing a reason to act that agents have themselves. It has no binding power over them, unlike their own desire for the goal that p.
The word ‘norm’ can also be used to mean “what an ought expresses” — the prescriptive counterpart to a (descriptive) “proposition”. Abstraction here works as a protective smokescreen. The many difficulties associated with the concept “proposition” are inherited and worsened.
Even if all members of a group act so as to bring it about that p, a new member A might have no reason to act in that way. No matter how many other similarities they may share, without a desire that p, there is nothing to bind A to act so as to bring it about that p.
The word ‘normative’, derived from ‘norm’ and intended to mean roughly “having the binding force of an ought” is also misleading inasmuch is it stays true to its etymology — suggesting as it does that a rule or standard is required for binding power.
It is sometimes said that “belief is normative”. That sounds like a category error, as a claim might be normative (by being an ought) but a belief is a thing, not a claim. Beliefs have content, like claims, but contents are usually descriptive rather than prescriptive.
Perhaps the best that can be made of the claim that belief is normative is this: beliefs play an indispensable part in fixing the binding power of oughts. And we are guided by oughts when we attribute beliefs to agents.
The latter is no minor matter, as beliefs just are what correct attributers of belief say they are, guided as they are by principles of charity. As Davidson noted, there’s nothing more to learn about belief than what a fully-informed interpreter would ascribe as content.
Principles of charity include: presumptions in favour of the truth of an agent’s beliefs and the coherence of his desires; a presumption in favour of his logical competence; and a determination to lower our expectations — to work “downwards” — only when necessary.
Such principles of charity guide all attributions of thought to others (and ourselves — “charity begins at home”). Because thoughts themselves are delineated and individuated by such attributions, we might call these principles “the oughts of thought”.
These principles of charity are not expressible as factual claims. They are oughts that guide conduct — the activity of interpretation. The essential ingredient of all oughts being goal-directedness, this puts goal-directedness or desire at the very centre of all mental life.
But we must be careful not to think of desire as a subjective experience of yearning. It is instead the cause of habitual behaviour aimed at a fixed goal. A strong desire is not a vivid sense of yearning, but fixation on a goal that overrules similar fixations on other goals.
In effect, to posit a goal for the purpose of prediction and explanation is to adopt Dennett’s “design stance”. To go further and also attribute representational states which co-vary in a reliable way with states of the world is to adopt his “intentional stance”.
In a rudimentary agent such as a thermometer, goal and representation might be “keep the room at 70°” and “the room is now at 65°”, the latter embodied by curvature of a bimetallic strip. These are rudimentary analogues of desire and belief, posited by the intentional stance.
The difference between these rudimentary states and full-fledged desire and belief is one of degree — of interconnectedness. An agent with genuine beliefs has a belief system, a detailed map of his world. An agent with genuine desires has many goals, a blueprint for his world.
We can be assured of the reality of these states: they are reasonably determinate because interpreters converge on their content, and routinely utter truths about them. And such states have causal powers, because causal links are trivially written into them.
For example, a belief that it is raining is typically caused by the fact that it is raining. A desire to have a cup of coffee typically causes an agent to buy a cup of coffee. The interpretative process that attributes contents like these presupposes such causal connections.
So mental states like beliefs and desires, posited by the intentional stance, are “woven by interpretation”. This has some interesting side effects. For example, an interpeter can’t assign too many false beliefs to an agent. The result is that “most beliefs have to be true”.
None of these reflections is intended to work as an argument against radical scepticism, as they assume a great deal about the world and the nature of mental representation — assumptions that the radical sceptic claims to forbid himself.
Because principles of charity embody rationality, they make it impossible to ascribe out-and-out irrationality, such as the having of contradictory beliefs. Agents can have odd preferences and false beliefs, but they can’t act “contrary to reason”.
That is because charity enjoins us to re-interpret the contents of an agent’s mental states in such a way as to admit that we were originally mistaken about them, and accept that the agent acted reasonably after all, even though his beliefs and desires may now seem bizarre.
That is not to say that agents can’t pretend to act unreasonably. In the traditional version of vranyo, an agent blatantly utters known falsehoods as a show of strength; an agent can also deliberately act “unreasonably” as a show of strength.
It can work as a show of strength because falsity and unreasonableness usually “set you back a bit”. If you can afford it, you must be well off. If you can thrive, you must be rich indeed. To reject obvious truths or justice is to set oneself back by defying social expectations.
In effect, that is to trumpet one’s place at the top of the pecking order.
(Petruchio: “I say it is the moon.”
Katherine: “I know it is the moon.”
Petruchio: “Then you lie. It is the blessèd sun.”
Katherine: “Then God be blest, it is the blessèd sun.”)
We routinely attribute mental contents to animals like dogs as well as to humans. But such attributions and the states they individuate correspond to specifically human vehicles of communication (declarative sentences) and human categories (words of our language).
There is a disconnect between actual thought processes of the brain (with its activation spaces and activation vectors) and our attributions of mental content (which are shaped by the needs of communication). The former are typically continuous, the latter discrete.
Holism — the idea of the twentieth century — has done much to bridge this gap. The future may bring new ways of thinking about knowledge — no longer as “justified, true belief” but as reliable, faithful representation of non-sentential form.