Max van Duijn
The lazy mindreader : a humanities perspective on mindreading and multiple-order intentionality
(2016)
[9]
Introduction
...
[13]
...why...The Lazy Mindreader? ...the short answer is this: we humans live in a socio-cultural environment that allows us to be “lazy” regarding the investment of mindreading efforts most of the time. This environment, of which the conventions underlying language and interaction are an important part, contains the coagulated experience of many generations interacting with each other and the world around them. Globally, I think that most approaches to mindreading have placed too much of a burden on individuals as “isolated cognitive units”, and paid too little attention to the ways in which this burden can be alleviated by, for instance, lexical items, grammatical patterns, or narrative strategies, and by the interlocutors (including writers/narrators) we cooperate with to make interaction work. My alternative view focuses on economy and least effort: processing of complex networks of intentional states is not seen as something the lazy mindreaders envisaged in this thesis do by default, but rather as a skill that is needed when the context requires deviation from a default—and even then, I will argue, is these mindreaders’ processing often supported by mechanisms that are part of, mediated by, or closely tied to language and narrative.
[23]
Chapter 1
...
[32]
In a proposition describing a particular state of affairs in the world, it is usually possible to substitute words with other words that refer to the same entity without consequences for the truth-value (or even referential value) of the sentence. In other words: “this rule is simply the logical codification of the maxim that a rose by any other name would smell as sweet” (Dennett, 1983: 344). To give an example: provided that Macbeth and Hamlet were written by the same author, it should be possible to substitute “the author of Macbeth” with “the author of Hamlet” in a proposition, without the truth-value and referential value being affected. The propositions (9) and (10) are thus either both false or both true:
(9) The author of Macbeth was born in Stratford-upon-Avon
(10) The author of Hamlet was born in Stratford-upon-Avon
[33]
However, in the following two propositions this is not necessarily the case:
(11) John believes that the author of Macbeth was born in Stratford-upon-Avon
(12) John believes that the author of Hamlet was born in Stratford-upon-Avon
After all, what John does and does not believe is independent of the “real-world fact” that the author of both pieces is the same person. This is what Dennett calls referential opacity: “the terms in such clauses are shielded or insulated by a barrier to logical analysis, which normally “sees through” the terms to the world the terms are about” (1983: 345).
The take-home message from the substitution test is really that there is a relation of dependency between the intentional agent and the non-intentional proposition. In the case of “John believes that it is raining outside”, the intentional agent (John) and the intentional expression (believes that) are the responsibility of the speaker, the one who asserts the proposition, whereas the non-intentional proposition (it is raining outside) is placed under the responsibility of the staged intentional agent (John). As a consequence, it is “insulated”, in Dennett’s words, “shielded from logical analysis” (1983: 345).
...
[35]
...the non-transitive, dependent nature of the relationships exhibited by multiple-order intentionality propositions is one of the core aspects of the problem dealt with in this thesis.
1.1.3 The intentional stance So far in this thesis, different intentional relationships have been categorised using mentalistic expressions from everyday language, such as thinking, knowing, believing, desiring, intending, and so on. How can we be sure that these terms are appropriate? Do they correspond to the actual intentional states held by others around us? ... According to Dennett (1983; 1987) we do not need to be sure. He argues that in order to understand phenomena in the world, one can adopt various strategies or “stances”, corresponding to different levels of theorising (partly overlapping with Marr’s levels of explanation,...
...“that fox digs a hole because it wants to build a nest” or “bird X believes that bird Y is hiding food”. Dennett argues that usage of everyday language is not problematic in such cases, as long as one keeps to the appropriate level of theorising. ... one can perfectly well make use of everyday mentalistic vocabulary as long as one is dealing with questions of some beings’ behaviour in their social environments, and not with the “lower level” mechanisms and physical processes underlying social living.
Now we're getting somewhere! But for artists, I think, the lower level is not the concern. Rather, we have to worry about an "upper" level, a sort of social "superstructure" of which observable behavior is the "base."
i.e. When the social environment begins to demand that an artist's intention bear some significant social weight, this is precisely where and when the razor's-edge failing of an everyday mentalistic vocabulary is suddenly felt more acutely.
[51]1.2.2 The roles of language
...
[54]
...Levinson (2006) introduces the concept “Schelling mirror world”. Schelling was an economist who studied a specific species of coordination problems: the ability of subjects to arrive at a solution together in the absence of
[55]
communication. For example, if they are told that they have to meet someone else in Moscow the next day, but not exactly where and when, and they know that the other has had precisely the same instruction, they can perform much better than chance would permit by (implicitly or explicitly) asking themselves what the other will think, and what the other will think that they will think. A “Schelling point” (Schelling, 1960) high above the odds in Moscow is probably “12 noon at the Red Square, in front of the clock tower besides the Kremlin”. If one has to meet in a theme park, this point would probably be the entrance, or in a crowded department store it may be the “lost-and-found” desk. Converging on such Schelling points, according to Levinson, requires not only a special way of reflexive thinking (about what the other will think one will think, etcetera), but also a notion of mutual knowledge or common ground, including a sense of mutual salience: “what leaps out of the common ground as a solution likely to independently catch our joint attention”... He argues that these same ingredients are also requirements for human communication: reflexive thinking and common ground, including a mutual sense of salience. After all, as has been described by many linguists and philosophers of language, there are thousands of possible ways in which a particular meaning can be expressed, while at the same time, every expression can have many different meanings. Only through the same combination of reflexive thinking and common ground, including a sense of mutual salience, can humans coordinate their mindstates while interacting, or in Levinson’s words: it is through these factors that “meetings of the mind” can occur in the “Schelling mirror world” that underlies human interaction...
...
[56]
...
S intends
that A should recognise
that S intends
that A should believe
that z
...instead of suggesting (following Sperber and Scott-Phillips) that interaction works because interlocutors (somehow, implicitly) take the steps spelled out above in order to “meet” each other at five orders of embedded intentionality, I argue that as a rule they start off having already met—and instead of suggesting that it is necessary by default, I suggest that it is only in exceptional cases that such steps need to be taken...
Put differently: in theory it is possible for interlocutors to reflect on the communicative situation in the way suggested by Sperber and Scott-Phillips, but in practice it is rarely necessary. Normally, a signaller “tosses” a particular behaviour (typically a string of sounds, gestures, and facial expressions) into the Schelling mirror world, assuming that the addressee will be able to figure out what the signaller means by it.
...
[57]
...
in interaction we do not have to sort everything out by default and right away—in every communicative turn we seem to build representations that are “good enough” for the interaction to keep going, but no better... If required, interlocutors can work out a particular point in more detail, aiding and steering each other in the desired direction turn by turn. Many conversations do not have “signal-response” as their basic structure, but rather “testing-adjusting-retesting” (Levinson, 2006).
Well,
then when tf do
addressees
of
message cinema, e.g.,
ever
adjust
or
retest
?
If we're ultimately headed for the argument that
Learning to deal with narrative may...hone one’s “real-world” capabilities of switching between multiple perspectives, understanding situations in terms of the underlying perceptions, intentions, motives, etcetera, and mapping behavioural patterns on particular mental states.
(53)
then wtf to do when our ways-of-honing have themselves become always-already concretely communicative as opposed to merely suggestive?
i.e.
narrative may...
...but do it?
[77]
2.2 Mindreading, Theory of Mind, and Multiple-Order Intentionality ... In this article we will follow Apperly’s suggestion to drop the term ‘theory of mind’ (to avoid the implication that attributing mindstates is like having a theory) and refer to the set of mechanisms, routines, and tricks that humans apply to form understandings of other’s mindstates as ‘mindreading’.
Glad to finally find this addressed speficially. Frankly, mindreading seems every bit as problematic, if for different reasons.
One thing that's important about having a theory is that theories are prospective and provisional, i.e, they await proof. The above seems skittish about an unrelated aspect of "theory," which admittedly is in fact more on the surface of the word, i.e. its connotation of conscious, effortful deliberation. But when you don't live in this particular academic world, I think you could be forgiven for finding the conceit to "mindreading" of any kind to be a bit much. We can only have "theories" about other people's thoughts and feelings, no?
[80]
2.3 Representing embedded mindstates: from sentence to narrative ...if the representation of a MOI ["multiple-order intentionality"] situation relies on such complex sentences only, it very soon becomes hard or even impossible for a reader or hearer to make the right inferences about the involved mindstates. ... This is in line with the fact that in corpora of narrative fiction and natural spoken discourse, sentences containing three or more embedded mindstates are very infrequent.
At the same time it is widely accepted that we do regularly deal with higher-order mindreading in daily social life, which suggests that there must be ways to communicate about this linguistically.
I don't see why that is necessarily suggested . I thought that was the whole point of the "Schelling mirror world?"
In our example of Shakespeare’s Othello, situations that require higher-order mindreading only emerge gradually in the course of the plot,... Nowhere in the text does any single sentence express the embedding of more than three mindstates: only after a while does the situation give occasion to consider forming such sentences, summarising the state of affairs in terms of embedded mindstates. ...
...
... As Iago’s scheming plan progresses, keeping track of ‘who knows what’ (including what they know that others know, etc.) involves reasoning up to at least sixth-order intentionality,...
[82]
... However, as stated earlier: nowhere does the text contain any sentences describing more than three mindstates at a time. This is where narrative takes over.
So, narrative takes over when the web of intentionalities becomes too tangled.
Epistemologically, what are the prospects for simply avoiding these "higher orders"? Is there anything to be said for avoiding altogether the "narrative" construal of "intentions?" As a thought exercise, let's take advantage of having been presented the notion here and presumptuously elevate this distinction between the Theory way and the Narrative way to a guiding epistemological principle. Let's draw a/this line in the/this sand so as to well and truly create a righteous distinction between those who "think in story" and those who don't.
Not even the American jury trial goes quite this far, it is true, but the trial does have a peculiar way of working through things, incrementally litigating small "truth claims" and then "building a case" out of those which validate. Yes, "intentions" are implicated in some cases; no, intent is not established by looking into someone's eyes. If we perhaps put too much faith in legal formalism given its well-documented capacity to miscarry, we are clearly uncomfortable reverting to folk wisdom once the offense surpasses a certain degree of severity.
Needless to say as well, if this paper is at all in the ballpark, then it's more obvious than ever why Othello and everything else I read in High School English failed to make much of an impression. I am not someone who is particularly drawn in by "sixth-order embedded intentionality." Frankly that sounds like some inner circle of yenta hell. But I would dispute (vehemently) that this betokens some kind of deficiency. It's more like an over- rather than under-sensitivity to social stimulation.
[96]2.5.2 Context and directions
... It can take up to the age of 12 or 14 before an adult level of understanding other minds has been achieved. It has been suggested that the development of the second layer is to a great extent influenced by social interaction and increasing experience with communicative situations, but also by language acquisition as such. ...
[97]
...Whereas existing research in this domain has focused on the developmental step from no or very basic understanding of intentionality at age 1–2, to the full appreciation of (false) belief-states at age 4–5, it would be of particular interest to trace the steps beyond that point: those involving higher-order mindreading.
In this article, we focused on the possibility of employing insights from rich traditions in the humanities, such as narratology and the study of grammar and semantics, to contribute to current debates concerning human cognition. This demonstrates how research on such multifaceted topics as mindreading can benefit from the integration of insights and methods from different academic disciplines across both the sciences and the humanities – an example of how EO Wilson’s idea of consilience (1998) can be brought into practice.
Hmm. Sounds more like the opposite.
[99]Chapter 3
Not afraid of Virginia Woolf: embedding and polyphony in
the novel
...
[102]
3.2.1 An early cognitive literary scholar
...
[103]
... It is a characteristic of modernist fiction more generally to be fascinated by people’s inner lives (Korsten, 2005: 227-249) and when reading what Virginia Woolf had to say about this herself one is easily tempted to call her an early cognitive literary scholar.
[104]
My first assertion is one that I think you will grant—that every one in this room is a judge of character. Indeed it would be impossible to live for a year without disaster unless one practised character-reading and had some skill in the art. Our marriages, our friendships depend on it; our business largely depends on it; every day questions arise which can only be solved by its help.
[…] But it is the art of the young. In middle age and in old age the art is practised mostly for its uses, and friendships and other adventures and experiments in the art of reading character are seldom made. But novelists differ from the rest of the world because they do not cease to be interested in character when they have learnt enough about it for practical purposes. They go a step further; they feel that there is something permanently interesting in character in itself. When all the practical business of life has been discharged, there is something about people which continues to seem to them of overwhelming importance […] And this I find is very difficult to explain: […] what the impulse is that urges them so powerfully every now and then to embody their view in writing.
[…] I believe that all novels, that is to say, deal with character, and that it is to express character—not to preach doctrines, sing songs, or celebrate the glories of the British Empire, that the form of the novel, so clumsy, verbose, and undramatic, so rich, elastic, and alive, has been evolved. (Woolf, 1924: 2-11)
Well, this peculiar expression of character indeed ramifies into a doctrine as soon as such a statement as this one has been issued, no?
The first skeptical question, again, is: why can't this "character" be resolved into observable behavior? Why attempt to "mindread" (even lazily) when you can (even more lazily) sit back, observe, and take note?
Part of the answer, apparently, is that here we have already aspired to predict behavior before it manifests. Evolutionism and selection pressures are called in to show that we cannot, after all, merely avoid making such predictions (or not if we want to survive). But the caveat that selection pressure drives not towards maximum but rather minimum fitness seems especially important here. Human beings are worse at few things than predicting the future. Here we are admonished to believe that we're actually pretty good at it in a few very specific applications. But I wonder what evolutionary theory by itself actually says about our abstract good-ness at this task? It seems merely to say that we must have improved over what our less successful primate ancestors were able to do. But bare survival, the ultimate test, is actually no test at all of the particular conceits attaching to the sophisticated literary thought of a postindustrial society. To evaluate those sorts of pretensions, we need finely targeted and controlled investigations.
Another part of the answer: character is resolved into behavior in quite a direct way some of the time, namely when it exceeds a certain degree of severity. Remember Taleb: no one considers a plane crash to be "anecdotal." A single plane crash becomes an empirical data point for all sorts of prospective epistemological purposes, but a single fender bender is unlikely to matter much in this way. Of course this line of thinking may just as easily suggest movement away from the argument I want to build here: a murder committed under unusual circumstances, as a "behavioral" data point, may not actually square very well or at all with the person's overall "character," which may otherwise (seemingly) be better-than-average. People of course cannot be totally good or totally bad, and occasionally they can be "complicated" to an incomprehensible degree. Even a murder, which indeed "no one considers anecdotal," may be misleading in precisely this way. Enter the narrativizers! But if that is the state of things, then what exactly are we chasing in the first place with this need (which I accept!) to establish the "character" of our cohort? It seems there is a danger of coming to enjoy the pursuit a little too much, perhaps of making things harder than they need to be; e.g.,
Dunbar also argues that people take a certain delight in being pushed to the boundaries of their cognitive limits, and that it may be a hallmark of literature to do so.
(101)
I don't know how to evaluate that contention, but there is of course the passage from Kahneman:
the current interview had failed at least in part because it allowed the interviewers to do what they found most interesting, which was to learn about the dynamics of the interviewee's mental life. Instead, we should use the limited time at our disposal to obtain as much specific information as possible about the interviewee's life in his normal environment.
(Thinking Fast and Slow, 230)
I know not whether this shift of focus from "mental life" to "life in his normal environment" in fact betokens a greater focus on oberservable behavior, but it does (explicitly) indicate less focus on la vie interieure. And it suggests that the latter pretension was not very predictive when applied to a mundane empirical question.
[105]3.2.2 Zunshine’s “sociocognitive complexity” ...I argue with Zunshine that the passage from Mrs Dalloway...suits an analysis in terms of mindreading well. However,...I will question three assumptions...: (i) that this passage (or the novel more generally) confronts readers with a form of complexity adequately conceptualised by counting layers of embedded mindstates; (ii) that these embedded layers pose a highly demanding cognitive processing task at the limit of these readers’ abilities; and (iii) that their appreciation of the literary work is somehow affected by this layered complexity. ...
[106]
... She suggests that different genres may implicitly expect different levels of “sociocognitive literacy” (i.e. aptitude to deal with embedded mental states) from their readers: for example, Jane Austen’s Pride and Prejudice may in its original form expect readers to deal with fourth- or fifth-order intentionality, whereas the same story in a comic-book adaptation or study-guide synopsis might be “downgraded” to second order, thus anticipating lower sociocognitive literacy... In addition, in several publications Zunshine links the levels of intentionality to perceived literary quality, arguing that there may be a “literary sweetspot” at which the number of embedded layers is optimal:...
Zunshine’s approach of analysing literary texts in terms of levels of embedded intentionality has found its way into the work of other scholars across both the humanities and sciences.
They must have been absolutely overjoyed to discover it.
[125]
3.4 Discussion and concluding remarks ...
[126]
A perspective on the analysis offered in this chapter...follows from the work of Bakhtin. One of his
[127]
central points is that characters and their perspectives in a novel should be seen as “dialogising voices”, interacting also beyond what is literally written down in the text... ...though an author or narrator can give her best attempt to “orchestrate” their voices, she will never be able to fully control them... One important reason for this, according to Bakhtin and his adherents, is that readers will always, consciously or unconsciously, attribute all kinds of thoughts and intentions to characters. It is for sure thinkable that a text can bring a reader in a position where it is both hard to access a particular character’s thoughts and necessary to do so in order to be able to understand the story—in that case, it is also thinkable that reading would be cognitively demanding. However, as far as the literary excerpts analysed in this chapter are concerned, quite the opposite situation seems to obtain: texts are full of elements that make the contents of different mindstates and their mutual relations accessible instead of opaque, and, thinking in the way proposed by Bakhtin, readers seem prone to (over)attributing intentional relationships when interacting with the text easily rather than by virtue of hard cognitive effort.
...[the assumption] that readers’ appreciation of a literary work is affected by the amount of mindreading complexity it contains. Rather than being drawn to fictional texts because it is so hard to figure out what all the characters are thinking..., we might just as well conclude that fiction attract us because it takes us relatively little effort to get access to rich representation of others’ inner lives, compared to real-life settings where we usually do not get such “360-degree spectator sight”. ...[perhaps] we are confronted with a rich thoughtscape, constituted by a polyphony of different voices,... ...[but] unlike in our daily lives, in Mrs Dalloway the text provides us with a full panoramic tour of all the mental space we normally don’t get to access so easily. Put differently: as readers of fiction such as Woolf’s, we can sit back in our chairs lazily and enjoy the thoughtscape, while the narrator does most of the mindreading for us.
A welcome tables-turning which seems applicable to just about any/all of the more-is-more school of academic "humanities" jargonmongering.
[129]Chapter 4
Viewpoint Packages: linguistic tools for communicating and processing complex “thoughtscapes”
...
4.1 Introduction ...
...the choice between accident or murder coincides completely with the construal of the athlete’s mindstate at a particular moment during the night of the shooting. Ultimately, this construal was made by a judge in court, based on information from forensic research, interrogations, witness reports, and so on. But besides that, and from the very first day after the incident, thousands of people have made such construals for themselves, mostly relying on cues presented by the news media.
So we will actually be considering a (potential) murder case here! Will we also be considering the political aspects of making such construals for ourselves as against forensic research and witness reports ?
i.e.
Was not a revolution once set ablaze against despotic magistrates and vigilantes who fell a bit too deep in love with their own
construals
?
[134]4.3 Constructing complexity
4.3.1 Compositional complexity: a literary example
...
[135]
...
Construction and elaboration of narrative spaces take mostly place through processes generally known as framing...and blending... The basic idea is that the background knowledge used by readers to interpret a text is structured in frames. Particular linguistic items used in the text activate a frame in its entirety, even if they relate to it only indirectly and in an unpredictable manner... Using this principle, writers of news articles can evoke rich meanings while providing only a few cues. For example, if a headline provides the information that a neighbour heard “non-stop shouting” coming from the Pistorius home during the hours before the shooting, nothing more needs to be said to evoke the frame of a fight between lovers,
...
[201]
Chapter 6
The Mentalising Test Revisited
...
[207]
6.2 Testing mentalising competence
...
[211]
6.2.1 Narratives and propositions
...
[213]
...
...propositions of the form used for the higher-order questions [e.g. "Sam believed that Pete thought the Post Office was in Elm Street and hence that Pete must not have intended to mislead Sam"] do not exist in the “wild”, so language users cannot rely on experience when assessing
[214]
them. The stories communicate the same information in a somewhat more natural way: in fact, they offer another demonstration that in natural communication “narrative takes over” when more than three perspectives are involved. ...
issue (i) Classic mentalising tests use narratives to present a social situation, but use propositions to present the questions. Especially above complexity level 3, such propositions are a very unnatural way of representing intentional states in discourse. With the analysis from Chapter 2 in mind, I suggest that the unnatural presentation of the questions can be a factor limiting performance on especially higher levels of complexity.
...
[215]
...
issue (ii) Narrative language usage features all kinds of cues that prompt and mutually coordinate intentional states of characters. As the analysis in Chapter 3 showed, a thoughtscape emerging in this way is easily misrepresented by propositions featuring only embedded mindstates. This suggests a structural discrepancy between the nature of the complexity presented in the stories and in the questions.
...
[219]
6.2.2 Packages and inferences drawn from the common ground ...
[221]
...
issue (iii) Following the same line of reasoning leads sometimes to a correct and sometimes to an incorrect answer, suggesting that factors other than the amount of orders of intentionality involved in a question...can quite easily interfere with error rates produced by participants.
...
[224]
...
issue (iv) Participants are asked to reflect, in the same way, on:
- viewpoint layers that are relevant to the characters and the development of the plot (whether or not these are spelled out compositionally or cued holistically); and
- viewpoint layers that can in principle be inferred from the story, but do not have such relevance....
6.2.3 Judging facts and intentional states ...
[225]
...
issue (v) Questions exhibiting unbroken chains of embedding have to be processed as a whole, whereas in questions containing conjunctions and/or causal links each constituting element can be checked against the story separately. Given that the first category contains only (some of the) intentionality questions and the latter all factual memory questions (and the rest of the intentionality ones), this may have affected the observed difference in performance on both types of questions.
...
[230]
6.2.4 True and false statements
...
[232]
...
issue (vi) Some questions exhibit a disproportionate increase in complexity per level for questions where the correct answer is “true” versus those where the correct answer is “false”: if the correct answer is “true” participants have to check every element for consistency with the story, whereas questions where the correct answer is “false” can be answered by spotting one element that does not fit. This issue seems to be best avoided in questions that exhibit an unbroken chain of embedding and that present
[233]
a scenario that is (about) equally credible when thought of as true or false.
issue (vii) Given issue (vi), there is a crucial difference between questionnaires using a true/false design and those using a forced-choice design: when using the latter, spotting a single “false bell” is possible in all questions (although it is more difficult in questions exhibiting an unbroken chain of embedding and presenting two equally credible answer options).
No comments:
Post a Comment