I M a Bit Confused by the Distinction Between Commitment-Preserving

9/24/2012

Week 5 Questions

CCDP

I’m a bit confused by the distinction between commitment-preserving

inferences and entitlement preserving inferences, as spelled out in CCDP

pp. 12-13. It sounds like you are saying that commitment-preserving

inferences from A to B are ones in which the inference from A to B is

non-defeasible, and entitlement preserving inferences from A to B are those

in which the inference is defeasible. Thus one is not committed to there

being good weather upon seeing a red sky at night because the inference

could be defeated by the observation of a barometer, for instance. Is the

inference from X being a zebra to X having stripes commitment-preserving or

entitlement-preserving? The inference seems defeasible, as in the case

where X is albino, so I’m tempted to take it to be entitlement preserving.

If that’s right, commitment-preserving inferences sound an awful lot like

formal deductive inferences, especially as your example on p. 13 seems to

run:

(∀x)((x is a yacht . x is in the harbor) ⊃ (x is a sloop))

John B is a yacht . John B is in the harbor

Therefore, John B is a sloop

I don’t see why we need to take this to be a *materially *good inference at

all, rather than a formally good one. Furthermore, if the inference from X

being a zebra to X having stripes is simply entitlement-preserving, I fail

to see how we commit ourselves to anything we haven’t made explicit in our

reasoning process, except for, possibly, the purely analytic statements

which follow from what we’ve stated. On the other hand, if the inference

from X being a zebra to X having stripes is consequence-preserving, it

seems like it’s going to be quite difficult to draw a hard line between

commitment-preserving and entitlement-preserving inferences.

*MIE*

In p. 185, you mention that linguistic scorekeeping practice is “doubly

perspectival”, in that “What C is committed to according to A may be quite

different, not only from what D is committed to according to A, but also

from what C is committed to according to B.” I was wondering how essential

this latter perspectival aspect is. If A and B were to “make explicit” what

they attributed to C with the use of scare-quotes and de re ascriptions –

thus enabling them to make distinctions between C’s doxastic commitments

andsubstitutional commitments-- would they eventually always come into

agreement concerning C’s commitments? If the answer is no – is it possible

for A and B to both be, in some sense, “right” about C here, or will there

always be a fact of the matter?

*BSD*

`I’m a bit confused about the second component on the argument on pg 82:

Only something that can *talk* can do that, since one cannot *ignore* what

one cannot attend to (a PP-necessity claim), and for many complex

relational properties, only those with access to the combinational

productive resources of a *language* can pick them out and respond

differentially to them. No non-linguistic creature can be concerned with

fridgeons and old-Provo eye colors.

I was thinking that the difficulty you had in mind was no the ability to

ignore certain complex relational properties when they were irrelevant, but

the ability to decide *what *to ignore. How does having the ability to

deploy a discursive vocabulary aid us in this decision? Is it primarily

normative considerations?

Billy Eck

*MIE**, Ch. 3*:

Imagine an objector who claims that your deontic scorekeeping story of

discursive practice is insufficient for conferring meanings on to

assertions. She asks us to imagine two autonomous discursive practices

that (currently) share the same exact set of possible assertions (the

sentences *p*, *q*, *r*, etc.). In both ADPs, she stipulates, a speaker is

entitled to assert a sentence in the same circumstances. Both ADPs warrant

the same set of material inferences from one sentence to another. In

short, the scorekeeping is officiated in the same way in both.

In your story, she claims, the assertion of any sentence in either ADP must

be thought of as expressing the same content as in the other. But she

thinks that they don’t have to. For in ADP1, *p* means “There’s a rabbit!”

and *q* means “That’s blue!”, while in ADP2, *p *means “There are some

undetached rabbit parts!” and *q* means “That’s grue!” Your story, she

therefore thinks, is insufficient to confer meanings on assertions.

BB: Appendix II to MIE 6 on ‘gavagai’

Where has she gone wrong in her reasoning?

*BSD**, Ch. 3*:

This is more of a drawn out line of thought than a question, as (I think)

I answered what was to be my initial question. I'd like to send it anyway,

as I think bringing it up in class might shed light on the AI debate before

*BSD*. I'm also interested in whether you think I made some error in the

discussion.

In the section on arguments against AI functionalism, you paraphrase a

locus of Hubert Dreyfus’s criticisms of classical symbolic AI and translate

it into MUR-talk to make it address an “algorithmic pragmatic elaboration”

conception of AI (78, 75).

Dreyfus’s criticism is, roughly, that classical symbolic AI requires that

the ordinary practical skills that are necessary for our understanding and

going about in the world be codifiable in explicit rules within a program,

and that that reveals an incoherent intellectualism lurking in the

background that seeks to explain all knowing-how in terms of knowing-that. You

explain that “the corresponding argument against the substantive practical

algorithmic decomposability version of AI would have to offer reasons for

pessimism about the possibility of algorithmically resolving essential

discursive knowing- (or believing-)*that* without remainder into

non-discursive forms of knowing-*how*” (78).

I’m interested in what such an argument could look like and whether or not

it could count as having the form of argument that you later designate as

conclusive against the “pragmatic” conception of AI, namely, that “some

aspect exhibited by all autonomous discursive practices…is not

algorithmically decomposable into non-discursive practices-or-abilities”

(79).

Here’s a familiar suggestion from Gilbert Ryle’s classic chapter “Knowing

How and Knowing That”:

The consideration of propositions is itself an operation the execution of

which can be more or less intelligent, less or more stupid. But if, for any

operation to be intelligently executed, a prior theoretical operation had

first to be performed and performed intelligently, it would be a logical

impossibility for anyone ever to break into the circle.

p.30

Ryle takes this line of thought as conclusive against what he sees as an

intellectualism. There is always a knowing-how behind a knowing-that, not

vice versa. But does it have any bearing on AI? It says that

understanding a bit language will require some skill that is not itself the

understanding some other bit of language. But is reacting to the syntax of

some line of code understanding a bit of language? I’d presume no. So

Ryle’s point doesn’t seem to have much bearing on the AI debate. Dreyfus

seems to miss this when he takes his considerations to warrant pessimism

about AI.

But perhaps we can see why such a view is tempting, by appreciating your

claim that computer languages are in principle pragmatic metavocabulary for

some ADP. Because they play this role, it is easy to treat computers as

listening to the coding language so as to grasp what to do. But that’s the

wrong model. The computer language is not just a metavocabulary but also

constitutes the set of stimuli to which the computer, *qua* transducing

automaton, responds without having to understand what the syntax of the

stimuli might be used to express.

BB: Dennett on stupider and stupider homunculi.

Chuck Goldhaber

*Between Saying and Doing #1*

On page 46 of BSD, Brandom falls into a dilemma concerning how to

characterize his LX diagram. As Brandom has it, Vconditionals is VP-suff

forPinferring. Either inferring is understood on the act-object model or

not. Assuming that inferring is understood on the act-object model, any

case of inferring contains an act of inferring something from something,

and, independent from the act, what was inferred from what. On this model,

conditionals seem to state a relation between the objects of the act, not

on the acts themselves. As Brandom himself adds in a footnote to the LX

diagram, conditionals assert “explicitly that one thing that can be said

follows from another thing that can be said”, not that “the act of

inferring is permissible”. This makes it seem as if conditionals target

explication on relations among the objects of inference. Since the objects

of inference must be vocabulary expressing propositional contents,

Vconditionals and Pinferrings stand in a VV-suff relation, and

“Pinferrings” must then be written as “P*V*inferrings”. In the main text,

Brandom also claims that “what the conditional says explicitly is what one

endorsed implicitly by doing what one did”. This characterization of what

the conditionals explicate differs from the one in the footnote in that it

seems to make essential reference to a *doing*. However, this is of no

help to Brandom since the conditional makes explicit an *endorsement*, not

the ‘doing’ it was caught up in. On the current model, since endorsements

admit of an act and an object, the conditional makes explicit the relation

of the objects of the act, i.e., the relation holding between propositional

contents.

I can imagine a response according to which Vconditionals explicate only

those relationships between propositional contents that have been (in some

way) involved in the act of inferring; for how else are conditionals to

know which propositional contents stand in the desired relationship. This

would be in effect to throw off the act-object model as a fictional

idealization. Propositional contents are unintelligible apart from the

acts they are caught up in. But, this rejection requires that conditionals

themselves can be made sense of only by asserting one. So, “Vconditional”

must be written as “*P*Vconditional”. So, PVconditional now stand in a

PVPV-suff relationship with PVinferrings. Either take the act-object

distinction seriously with respect to inferrings, as Brandom seems to do,

and be left with a VV-suff relation, or else treat it as a fiction, but be

left with a PVPV-suff relation.

*Between Saying and Doing #2*

Brandom draws a distinction between algorithmic elaboration and practical

elaboration by training. That is a fine distinction, but it leaves out the

most intriguing aspect of most of our actions. We can act effectively and

in the appropriate manner in the face of novel and unexpected circumstances

for which we have not been trained. Brandom is perhaps right in

characterizing training as a sort of ‘feedback loop’ of perception,

responsive performance, and perception of the results of the performance. But,

a single run through Brandom’s loop would not adequately capture what is

unique about the sort of case I have in mind. This is because there was no

training for the act in those circumstances, the particular act was not

itself part of a training schedule, and that particular type of act may

never be performed again. Brandom’s loop would not show how such acts

differ in kind from practically elaborated ones. When Rossi passes Lorenzo

on turn 3 at Catalunyna (choose an example that works for you), he may have

not trained for the way he had to do it, nor was he training while he

passed Lorenzo, and indeed, he may never perform that way again. It is

obvious that no addition of pragmatically elaborated sub-performances

captures what is unique about such acts, since the question would be left

over of why the sub-performances were added up in *those* circumstances.

The apt characterization of such performance should, in the end, be given

in terms of the agent’s or performance’s* goodness *with respect to what

the agent is trained to do or the performance’s kind. The appropriate

characterization of Rossi’s action is the attribution of an *accolade* to

Rossi, “Wow, Rossi is great (as a GP racer)!”, or to the performance

itself, “What a close-off!” Because no training was involved in the

performance under those circumstances, we have nowhere else to look for an

account of the act except the agent or performance itself. Similar points

seem to hold for interpretation, the difference being that we are all good

at it, so no accolades are given.

*Making It Explicit*

Last time I asked about how malapropisms apply to a semantically backed

pragmatics. I suggested it presented a counter-example to your view. In

your response, you noted that one also must make pragmatic inferences in

order to understand someone, though no general theory could be given about

most pragmatic inferences we would have to make. In response, I want to

suggest that *no* inferring is required in understanding another person’s

words. I can set it out in a double dilemma. The inference is either an

act or not. If the inference is an act, it can be made explicitly or

implicitly. Of course, we do not need to explicitly infer when we

interpret a person’s words; if that were so, we would not need Brandom’s

book to tell us that is how we do it. I would know that is what I did in

the way I know other things about myself. So, perhaps, as Brandom seems to

suggest, we may infer implicitly in attributing commitments and

entitlements to others. But how should we characterize implicit

inference? Commitments

and entitlements are already propositional contents. And, in order for me

to infer one content from another, I must understand the propositional

contents involved in the inference. But, “I understand *p* implicitly”

makes no sense, at least when “understand” is considered an achievement

verb, as we are currently considering it. This might suggest that

inference is not to be considered as an act that goes on in interpreting

someone. Perhaps we are only *attributing *commitments and entitlements

that *already* have some requisite pragmatic inferential structure. But in

order to make room for malapropisms, we cannot rely on a structure already

in place. Therefore, it seems that in interpreting another person we need

not infer at all. (Much here depends on how we cash out ‘explict’,

‘implict’, and their adverbial counterparts.)

Another thing to tackle is whether we *attribute* and *take*: in what sense

are they doings?, what does it mean to do one of them?, are they merely

innocent philosophical idealizations?

*Conceptual Content and Discursive Practice*

Brandom claims that the ‘six consequential relations among commitments and

entitlements…confer inferentially articulated, and so genuinely conceptual

content on the expressions, performances, and statuses that have the right

kind of scorekeeping significances in those practices.” What does ‘confer’

mean? The Queen can confer knighthood onto someone, and in doing so the

person undergoes a shift in state from not being a knight to being a knight.

An analogous shift in state cannot be found in Brandom’s system. Commitment

and entitlements are commitments and entitlements to propositional

contents, but to be a propositional content essentially involves the

possibility of its expression in language. So, expressions are already

possessed with propositional content. The same can be shown to hold, I

think, for ‘performances’ and ‘statuses’. If this is so, what does

‘conferral’ mean exactly? This question is very similar to one I asked two

weeks ago.

Shivam Patel

BSD 3 introduces a novel kind of PP-sufficiency relation to be laid

alongside the (by then familiar) one on which elements of one practice must

be algorithmically re-arranged (in uniform and antecedently specifiable

ways of the kind an automaton can carry out) in order to yield another

practice. The new relation is called "practical elaboration by training".

Its job in BSD 3 is to provide some relief to those who fear that if

sapience turns out not to be algorithmically decomposable into primitive

(clearly non-linguistic) abilities, then it becomes wholly mysterious. This

worry is unwarranted, BSD 3 says, because there is this other,

naturalistically equally respectable, form of elaboration/decomposition.

My (probably simple) question concerns the relationship between these two

relations between practices. If two practices are related by the

training-relation (to speak in-officially), is there not generally good

reason to think that the algorithm-relation also holds? Sure, we have no

idea how it is properly spelled out. As BSD 3 affirms, the specifics "vary

wildly from case to case, and depend heavily on biological, sociological,

historical, psychological and biographical contingencies" (BSD p. 85). But

this does not show that there is not also a way of decomposing the target

practice algorithmically into the base practice. And it also does not show

anything about the a-prioiriknowability of _whether_ there is such a way.

Often, in philosophy, this is all we need, isn't it?

Matthias Kisselbach

MIE, chapter 3:

I would like to ask a clarifying question concerning the conceptual relation between the assertional/doxastic sort of commitment and the corresponding entitlement: Is the former concept of commitment can be understood solely in terms of the latter concept of entitlement?; In other words, is the former concept conceptually reducible to the latter?

On the one hand, Brandom explicitly denies a familiar way of achieving this conceptual reduction. He explicitly disapproves the traditional definition of the commitment (i.e., responsibility) in terms of the entitlement (i.e., authority) plus the formal negation, and vice versa (p. 160). (In his project, the commitment and the entitlement are rather used to define the material sort of negation, or incompatibility).

On the other hand, however, it appears (at least to me) that Brandom suggests another way to understand the commitment in terms of the entitlement all the way down. To begin with, according to his analysis (pp. 172-80), the pragmatic significance of assertion consists of two factors: (1) undertaking the commitment (i.e., responsibility) to show entitlement to it if challenged in an entitled way; (2) Unless this commitment is breached, entitling (i.e., authorizing) audience to the same assertion. Then, Brandom goes on to explain the commitment undertaken in (1) in terms of the entitlement that the audience can internally sanction the breaching subject by depriving her assertion of the power to entitling the audience to it. Certainly, the assertional/doxastic commitment essentially has its propositional content, and that content is explained partly in terms of the commitment-preserving inferential role it plays, that is, in terms of the other commitments it entails and is entailed by. However, given the basic pragmatist analysis of assertion above, each of these other commitments can again be understood in terms of the relevant entitlement all the way down. Clearly this story does not have a form of straightforward definition like the traditional one, but it still seems to me to be a sort of reduction of commitment to entitlement.