Metaphorical Attitudes John Barnden School of Computer Science University of Birmingham, UK Research supported by: Leverhulme Trust Go Figure, June 2013, London Themes of Talk • Form unprecedented(?) link between – philosophy of propositional attitude reports (PA reports) – research on metaphor about mental states, especially in Cognitive Linguistics …. • Centred on the meaning intention problem [cf. notably Schiffer, Soames, Richard, Clapp; also Barnden]: the danger of PA-report theories implausibly imputing abstruse (non-commonsensical) thoughts about PA holders’ mental states to ordinary people. – I will call it the abstruse imputation problem. • Proposal: impute instead thoroughly commonsensical (~ folk-psychological) thoughts about mental states to ordinary people, where those thoughts are often metaphor-based, using a variety of different conceptual metaphors. • Issues like de-dicto/de-re and nature of reference are partly metaphor-based and, moreover, relativized to specific metaphors. An Analogy to the Issue • Consider “Sirius has exploded” Harry [the hearer] Suppose we based the content of the sentence on a scientifically correct analysis of a star exploding, where that analysis was not something that an ordinary person could be expected to grasp---e.g., it might involve nuclear fusion, quantum theory, general relativity, … • We could not use this content as it stands as the content that Harry grasps. To do so would be to impute abstruse conceptions to Harry. • Even if the analysis were only slightly non-commonsensical it would be completely wrong to claim that that content was understood by Harry, if one were seriously attempting to give a true account of Harry’s understanding. • However, that does not mean there would be no role for that scientific content. To keep such a role, one might claim something like the following: An Analogy, contd The commonsense content actually understood by Harry is a disguised version of the scientific content. E.g.: There is a systematic way of accurately translating the scientific content into commonsense notions. The translation in effect involves obscuring the abstruseness of the scientific content by commonsense “modes of presentation” / ways of thinking. The translation MoPs up the abstruseness. Abstruse Imputation Danger • (S) “Ralph believes that the man at the door is a spy.” • Suppose have a theory of believing that brings in some relationship of Ralph to theoretical entities of one or more of the following types: – situations as in situation theory, propositions, sentences of mentalese, Fregean senses & concepts, modes of presentation (e.g., of the man to Ralph’s mind), ways of thinking (e.g., about the man, in Ralph’s mind), mental files (e.g. holding some info about the man), …………. – E.g., one takes the above to mean (as one possibility) that Ralph is in belief relation Bp to a proposition composed of such things as the door, spyness, and location, together with MoPs. • Entities of the above sorts are brought in or elaborated in certain ways in order to cope with issues of transparency or opacity of parts of PA report complements. (S) may be true if Ralph is thinking of the person as the man at the door, even though the man is Ortcutt but Ralph does not believe that Ortcutt, thinking of him via that name, is a spy. Abstruse Imputation Danger, contd • Then the danger is that one will take an ordinary hearer Harry understanding (S) to think of Ralph’s mental state in the above sort of way: Harry conceives of Ralph’s mental state in terms of theoretical entities such as above. – E.g., Harry takes Ralph to be in relation Bp to a proposition composed of … and MoPs. • But to the extent that the above theoretical entities are abstruse, one is imputing to Harry an abstruse thought about Ralph that no-one except certain philosophers are capable of having. – Do ordinary people like Harry think about people as being in any relationship to any object as abstract and formally structured as, say, a proposition (of whatever variety that has been proposed in Philosophy?) Do ordinary people know what MoPs or ways of thinking are, when they are as detailed and carefully defined enough and finely individuated enough to do the job of coping with problems of transparency and opacity in PA reports? Abstruse Imputation Danger, contd • Above comments are abstracted from intricate discussions by Schiffer, Richard, Soames, Clapp, etc. • But more could perhaps be made of the following point. • Even if one could devise commonsensical notions of say, propositions (and of ways of thinking of things), and of a believer as being in a relationship to such things, more is needed than to credit Harry with the ability to think of Ralph as being in a relationship to a whole proposition. • One needs to credit Harry with the ability to think about gappy versions of propositions (or other comparable constructs) and to be able to imagine, e.g., a whole proposition as being composed of partially unknown parts. Abstruseness from De Re • E.g., a style of de-re interpretation of (S) is that Ralph believes some proposition that …… is at the door, where …… is – (a) some unknown representation (MoP, way of thinking, …) and – (b) refers to the man who is actually at the door. • So, Harry has to be able to think about a partially specified proposition in that sort of compositional way, and to think about parts of it as referring to things in the world. • Analogous comments apply to theoretical approaches involving relationships of Ralph to other constructs such as mentalese sentences or situations … or Soames-style cognitive acts (I think). • But can poor Harry really think about propositions (etc.) in this complicated, abstract way? Abstruseness from De-Re: Reference • And if he can, is he using one of the complex notions of reference studied in Philosophy, involving e.g. baptisms and causal chains, or a theory of information?? A Variant Case of the A-I Issue • The abstruse-imputation issue is a problem with nesting, in that the hearer (Harry) has thoughts about the thoughts of the PA holder (Ralph) – and we may in general also need to interpose the speaker Sally’s thoughts: a Harry/Ralph nesting or a Harry/Sally/Ralph nesting. • But it helps to throw the issue into relief to consider explicit nesting within belief reports: (Sa) “Peter believes that Ralph believes that the man at the door is a spy.” We have the issue of the way in which Peter might be thinking about (Ralph’s) act/state of believing. One thing we don’t want is for Peter’s way of thinking about Ralph’s believing to be abstruse. • The issues surrounding inner “believes” in (Sa) are similar to those for “has exploded” in “Peter believes that Sirius has exploded.” Nesting, contd • The point is that the general case is a nesting of PAs, such as A believes that B hopes that C intends that D believe that …, where A can be a hearer of a PA report that start with agent B, but other situations are possible. • At every agent boundary A/B, B/C, etc. we have to ensure that our theory of PAs or PA reports does not impute any abstruse thought to any of the agents, and does not involve any agent in the sequence imputing to the next or any later agent any abstruse thought. (Imputation of abstruse imputation must itself be avoided.) • However, we as theorists can have abstruse thoughts about what it is for the outermost agent A to believe, etc. – I.e., there can be an abstruse “outermost theory”. • And the whole theory must be able to impute abstruse thoughts to some agents – PA philosophers, for example! How to Proceed? • There may be come cunning way of deploying abstruse theoretical constructs on lines that have been developed in PA theory, but disguising them through commonsensical MoPping up, so that we get a nesting-adequate PA theory that avoids abstruse imputation. • But this strategy is – Full of potential for leading to hugely elaborate accounts (e.g., commonsense MoPs serving to MoP up abstruse MoPs) – cf. Ptolemaic epicycles – Potentially flawed anyway (or subject to yet more complexity) because of a multiplicity-of-view issue below. How to Proceed? contd • Propose instead an in-your-face strategy that embraces evidence about how people may actually think, commonsensically, about mental states, as revealed by fields such as psychology and cognitive linguistics. Build a PA theory on this basis (except for the case of the PAs of outermost agents) rather than starting from an abstruse theory. Irrespective of how partial or muddled those commonsensical views are, we only impute commonsensical views to agents (non-innermost agents in nested PA situations). • It seems that an important case of such views are metaphorical ones, and inclusion of metaphorical views may even be unavoidable. • Allow a range of views to be available to agents, not just one. On to metaphor … • They kicked the idea around the room. • She grasped the idea with both hands. • Many ideas were floating around in his mind. • In her mind, she could picture him sitting at home reading. • She believed in some dark recess of her mind that her boyfriend was unfaithful. • The idea slowly came to the front of her mind. • She could only remember him through a fog. • Part of him wanted to go to the party, but another part knew he ought to work on his essay. • She said to herself that she had been very lucky. • “I’ve been very lucky,” she thought. • A tiny voice said in her ear that she had been very lucky. • Part of her was saying that she was very lucky. • It was engraved on his mind that he should never borrow money. • For/To him, road safety was more important than defeating terrorism. • In his view, road safety was more important … • In his mind, road safety… On to metaphor … • A variety of broad “metaphorical views” (or conceptual metaphors, in Lakoff’s terminology) are in play, such as: MIND AS A PHYSICAL REGION IDEAS AS PHYSICAL OBJECTS COGNIZING AS SEEING (IDEAS AS PICTURES) IDEAS AS UTTERANCES (INTERNAL or EXTERNAL) MIND AS CONTAINING A WORLD AGENT AS HAVING A HIS/HER OWN WORLD (“AGENT’S WORLD” view) PERSON AS HAVING PARTS THAT ARE PERSONS PERSEON AS HAVING INNER PERSONS Caution re Internal-Utterances Metaphor • While sometimes the implication may be that the person is experiencing a linguistic utterance, this is not necessary. This may be clearest if indirect speech is used of an inner person: “Part of her was saying that she should apply for the job.” The ascription of speech to the person-like part is just an aspect of the fiction of the person contains such parts. • We can also say: “When a dog thinks, “Where’s my bone?”, he immediately tries to find it.” • But even when there is an implication that, or the hearer simply assumes that, the agent actually experiences inner language, – There is no necessary assumption that displayed utterance is a verbatim rendition of the inner utterance, or even in the same language. Cognitive Nature of Metaphor • Relatively abstract matters such time, love, money, mathematics and mental states are often described in language with the aid of metaphor. • A big claim in CL especially: metaphor is a commonsense cognitive tool, not just a communicative tool. • When we think about things, especially abstract things, we may often be using internal metaphor about them … and perhaps we even have to do so – NB: It may be extremely difficult, and perhaps even sometimes impossible, to give a purely non-metaphorical paraphrase of metaphorical utterances, especially if the paraphrase is to be commonsensical. • What it amounts to to use metaphor in one’s mind has seen considerable dispute. • I will take it to mean: sometimes when we are thinking about some topic T, we are actively using corresponding aspects of a source subject-matter S. Cognitive Nature of Metaphor • E.g., in thinking about time, we may use corresponding aspects of space; in thinking about a life, we may use corresponding aspects of a journey. When thinking about Y’s mental state, we may think of Y’s mind as a physical space, and/or her ideas as physical objects, and/or her thoughts as internal utterances or pictures, or … • Thinking metaphorically about a topic T may involve having aspects of T explicitly in mind, with links to corresponding aspects of S. But it might instead involve just the aspects of S, at least for some period. The thoughts in S terms could be linked to the person’s actions and communications well enough for a translation in T terms not to be (always) necessary. This is an especially important point when the only way the person has of thinking of T is via metaphorical sources S! Cognitively Added Metaphor • A natural corollary of that cognitivist stance to metaphor is that an agent can spontaneously bring metaphor into play. In particular: One can be thinking about T metaphorically in terms of S without processing any incoming utterance. Understanding an utterance about T can involve mentally casting T in terms of S even if the utterance does not do so. • E.g., “The meeting has been set for a later time” Harry Harry’s understanding of this could involve an imagined movement along a metaphorical time-line. That metaphor is cognitively added. This could even be (one of) Harry’s core way of thinking about times of events. It need not be just a helpful add-on. It is even possible he never thinks about event times (especially changes of them) other than through such metaphor. Metaphor Multiplicity and Inaccuracy • There are often several different metaphorical views for a given subject matter. • Different metaphorical views for crime – as a beast, as a disease, etc. • Individual metaphorical descriptions may be partial and otherwise inaccurate. • Different aspects of a situation may be best described via different metaphorical views, and several different metaphorical views may be used in conjunction to think more completely about something. The Proposal • An ordinary person, when thinking about a PA of another agent, usually adopts some metaphorical view or views of the PA. • In particular, a hearer of a PA report, even when it is not overtly metaphorical, usually adopts some metaphorical view(s) of the mental state of the PA holder. • The particular view(s) chosen (viewing the PA content as a physical object, an internal utterance, a picture, etc.) depends on the predilections of the hearer, context, the hearer’s current cognitive purposes, the detailed nature of the report, spill-over from recent metaphor use, etc. Examples • Harry might in principle interpret “Ralph believes that Ortcutt is a spy” as follows, assuming that Ralph is thinking of Ortcutt at least in part via the name “Ortcutt”. [The following are English glosses of some sort of representation in Harry’s mind.] – Ralph is saying to himself, “Ortcutt is a spy.” [where the quoted sentence is actually in English] – Ralph is saying to himself, “Ortcutt est un espion.” [e.g., if Harry is French] – “Ortcutt is a spy” is written down in a file somewhere in Ralph’s mind. – Ralph’s head contains a (moving) mental picture of a man going around doing spyish things. Ralph mentally names this man “Ortcutt”. – In the world according to Ralph, Ortcutt is a spy. [such a world is like a fictional world created by a short story] Agent’s-World and Situational Views • Can commonsensically conceive of Ralph as believing in a situation, e.g. believing in the situation of the person called Ortcutt being a spy. • But what is a situation, commonsensically? • I provisionally propose that a situation is just a mini “agent’s world” I.e. a small world that is part of the world as it seems to an agent (Ralph). • So a commonsensical situational account is basically the same as an agent’s-world account. Mixing • Multiple views can be entertained at once, as already apparent above. No reason why Harry’s view of Ralph’s belief cannot involve Harry imputing all of, say: a world, a picture, and an internal utterance to Ralph. • Harry might view Ralph’s mind as a physical space, with some memory being a picture that’s hidden in some recess of the space, or as an utterance being made at the fringes of the space. We get the locational connotations of Mind as a Physical Region plus the (sometimes deficient) content-representing facilities of pictures or utterances. Non-Awareness of Metaphor • Harry may not consciously realize he/she is thinking in terms of some metaphor for mental states in understanding Ralph’s state. NB: Ordinary people are not generally aware of metaphoricity in communication, or about how their reasoning is being swayed by metaphor. • It is easy to miss that one might be using a metaphor in understanding “Ralph believes X.” This thought about Ralph can appear to one’s consciousness in, for instance, the internal utterance “Ralph believes X” itself, accompanied by a feeling of understanding without one realizing why one has that feeling. • It is only when asked to explicate what it means to say that Ralph believes X that the underlying metaphor(s) comes consciously to the fore. Additional Points • A person may impute particular metaphorical views to other people. E.g., in an A/B/C nesting, person A may impute to B a particular metaphorical view of C’s PA. Similarly for deeper nestings. • The proposal does not preclude the use of commonsensical nonmetaphorical views of believing, hoping, etc., but it is not clear what these might be, except that … Behavioural Approach • A person may adopt a (commonsensical) behavioural (and perhaps dispositional) view of someone’s PA, especially if it is a believing, or impute such a view to another person. • I.e., a person might in some circumstances interpret “Ralph believes that Ortcutt is a spy” to mean that Ralph is in some respect acting as if, or is inclined to act as if, Ortcutt was a spy. • Such a view does not involve metaphor, provided that “acting as if” and “inclined to act as if” do not themselves need to be metaphorically conceived. Nested Cases • “Peter believes that Ralph believes that Ortcutt is a spy.” One possibility is that Harry thinks of (a) Peter’s head as containing a picture of Ralph’s head as containing a picture of a man doing spyish things, and (b) Peter saying to himself “Ralph calls the man Ortcutt in his mind”. Another possibility is that Harry represents Peter as saying to himself, “Ralph is saying to himself that Ortcutt is a spy”. • But in principle, nestings can involve a mix of views across agents. In an A/B/C nesting, A might conceive of B via agent’s world, and take B’s world to contain C who is mentally uttering something to himself or has a mental picture of something. Or, e.g., A could take B to be mentally uttering something of form “In C’s world, …”. Nested Cases: Internal Utterances • INTERNAL-UTTERANCES has a special feature. • E.g., Harry could use Peter is saying to himself, “Ralph believes that Ortcutt is a spy”. That is, once we have introduced an linguistic utterance, anything that could go into a real utterance could be used. • This is especially useful for deeper nestings: Mary is saying to herself, “Peter believes that Ralph believes that Ortcutt …”. De-Re and Reference again • Commonsense objects such as (real) sentences and pictures can commonsensically have gaps in them and attached commonsensical reference links. • An ordinary person can imagine a partial actually-spoken sentence “… is a spy” and can imagine the missing part as naming or describing some particular person. • That ordinary person need have no coherent, careful, complete theory of naming or describing in order to do this! • So in a de-re interpretation of “Ralph believes that Ortcutt is a spy,” Harry can (a) represent Ralph as saying to himself, “… is a spy”, and (b) specify that whatever is in the gap names or describes the person Harry knows of as Ortcutt. De-Re and Reference again, contd • Or, Harry could – represent Ralph as having a particular partial picture in his head (for this Harry can think of the picture itself), – where that picture has a gap much as above, or has an outline figure. Harry specifies that the gap or outline figure depicts the person Harry knows as Ortcutt. • Again, Harry need have no coherent theory of depiction by pictures, and may not have read Nelson Goodman on this! De-Re and Reference again, contd • In an AGENT’S-WORLD or situational account, de-re is handled by Harry postulating that an entity within Ralph’s world/situation is a counterpart of an entity in the real world [= the world according to Harry] … De-Re and Reference again, contd • Thus, the particular de-re options may be biased by what commonsense views (usually metaphors) are in play … and the nature of de-re may actually differ across views. • Especially because what it is, commonsensically, for a picture-part to refer to something could be very different from what it is for a sentence-part to do so. • And these are different from the idea of counterparts across worlds. Ideas-as-Utts and Metalinguistic Approach • IDEAS AS UTTERANCES of course has strong relationship to the metalinguistic approach to PAs [e.g., Elgin 1985], whereby a believer is in a relation to some natural-language expression. • Many difficulties with that approach as a scientific one, but they just become a particular example of how a metaphor can be inaccurate, or they can be folded into the very usage of the metaphor. – E.g.: problems with metalinguistic approaches with dogs or non-English-speaking people get washed out in the metaphoricity of IDEAS AS UTTERANCES. It’s just part of the metaphoricity that a non-English-speaker or a dog is cast as having an internal English utterance or French utterance or whatever. Further Remarks / Future Work • One needs to deal with additional varieties of PA representation such as dese. • One needs to account for the array of such varieties at each agent boundary within a nesting, with all combinations in principle possible. E.g., in A/B/C, A may have a de-dicto representation of B’s de-re representation of C’s mental state. • This leads to huge complexity in a traditional formal account but is easier to handle metaphorically precisely because of the familiarity and crudeness of metaphor. Further Remarks / Future Work • The advocated approach grasps the nettle of the fact that many constructs used in philosophical theory of PAs are already highly metaphorical, such as mental files. … And makes this nettle the centrepiece of the garden! • We have the following correspondences between (some) technical PA approaches and metaphors for mental states: – Notional world AGENT’S-WORLD – Possible-world elaborated non-commonsensical form of AGENT’S WORLD – Metalinguistic IDEAS AS SPOKEN? (INTERNAL OR EXTERNAL) UTTERANCES – Mental files IDEAS AS WRITTEN INTERNAL UTTERANCES – Situation theory AGENT’S MINI-WORLD / SITUATION – Relation-to-proposition theories metaphorical notion of a fact as some sort of entity in the world? – Relation-to-mentalese-sentence theories IDEAS AS UTTERANCES (indirectly)? Final/Future, contd • Possible world theory needs separate treatment. In brief: • Traditional, non-enriched PW approaches are not susceptible to the type of abstruse imputation studied here … • … but suffer the cost of having a hugely impoverished account of PAs. Final/Future, contd • The fact that many PA reports in real discourse are overtly metaphorical already makes metaphor an important thing for PA-report theory to address (and is somewhat separate from the question of embedding metaphor in PA report complements – cf. Stern 2000, 2006). • Folk psychology has been much studied in philosophy, but studies seem not to have taken on board the huge literature in metaphor theory, particularly within Cognitive Linguistics (CL). Much attention in metaphor theory has been given to metaphor for mental states in particular. • But equally, CL has largely not taken on board philosophical discussion of folk psychology, or technical issues such as de-re/de-dicto .. And has not dealt much with nested PA situations. And the CL idea that thought can be intrinsically metaphorical, and some thought may have to be intrinsically metaphorical, needs further philosophical study. Final/Future, contd • What difference does the unconscious/conscious distinction (or spectrum) make? Is unconscious common-sense different from conscious commonsense? (NB: metaphors of mind can variously describe either conscious or unconscious states.) • And how metaphor-imbued is the notion of consciousness, and how intrinsically metaphorical is consciousness itself?? [Barnden 2005; cf. some limited aspects of Jaynes 1982] References/Bibliography ATT-Meta Project Databank of metaphors of mind: http://www.cs.bham.ac.uk/~jab/ATT-Meta/Databank/ Barnden, J.A. (1986). Imputations and explications: representational problems in treatments of propositional attitudes. Cognitive Science, 10 (3), pp.319-364. Barnden, J.A. (1989). Towards a paradigm shift in belief representation methodology. J. Experimental and Theoretical Artificial Intelligence, 1 (2), pp.133-161. Barnden, J.A. (2005). Metaphor, self-reflection and the nature of mind. In D. N. Davis (Ed.), Visions of Mind: Architectures for Cognition and Affect, pp.45-65. Hershey, PA: Info. Scis. Publishing, Idea Group Inc. Barnden, J.A. (2008). Metaphor and artificial intelligence: Why they matter to each other. In R.W. Gibbs, Jr. (Ed.), The Cambridge Handbook of Metaphor and Thought, 311-338. Cambridge, U.K.: Cambridge University Press. Barnden, John A. (2010). Metaphor and Metonymy: Making their connections more slippery. Cognitive Linguistics, 21(1): 1–34. Clapp, L. (2000). Beyond sense and reference: An alternative response to the problem of opacity. In K.M. Jaszczolt (Ed.), The Pragmatics of Propositional Attitude Reports, pp.43-75. Oxford: Elsevier. Elgin, C.Z. (1985). Translucent belief. J. Phil., 82 (2), pp.74-91. Goodman, N. (1968). Languages of Art: An Approach to a Theory of Symbols. Indianapolis: Bobbs-Merrill. Jaynes, J. (1982). The origin of consciousness in the breakdown of the bicameral mind. Boston, MA: Houghton Mifflin. Richard, M. (1990). Propositional attitudes: an essay on thoughts and how we ascribe them. Cambridge, U.K.: Cambridge University Press. Richard, M. (2012). How must propositions be? In Working Papers No.~12: Workshop on Propositions and Propositional Attitudes, pp.178-196. Ecole Normale Superieure, Institut Jean-Nicod, Paris, 6 April 2012. Also: background in form of chapter entitled Mental Files in Attitude Ascription, pp.28-36 in same vol. (CONTD) References/Bibliography Schiffer, S. (1987). Remnants of meaning. Cambridge, Mass.: M.I.T. Press. Schiffer, S.(1994). A paradox of meaning. Nous, 28 (3), pp.279-324. Schiffer, S.~(2000). Propositional attitudes in direct-reference semantics. In In K.M.~Jaszczolt (Ed.), The Pragmatics of Propositional Attitude Reports, pp.13-30. Oxford: Elsevier. Schiffer, S. (2003). The things we mean. Oxford: Oxford University Press. Soames, S. (2010). What is meaning? Princeton University Press. (Also later workshop papers.) Stern, J. (2000). Metaphor in context. Cambridge, MA: Bradford Books, MIT Press. Stern, J. (2006). Metaphor, literal, literalism. Mind and Language, 21 (3), pp.243-279. On to metaphor … • From a trashy supermarket romance magazine … – Suddenly I was having second thoughts. About us, I mean. Did I really want to get married and spend the rest of my life with Mick? Of course you do one small voice insisted. Are you quite sure about that? another nudged. So much was going on in my head, I couldn't sleep. – Sharon pulled herself out of her jeans, the words ``How could he? How could he?'' jumping about her wearied brain. Senseless, leaving her empty, cold, helpless. Another voice, angry and vindictive, shouted in her ear, ``Serves you right, you silly fool: play with fire and watch your life go up in flames.''[.] My Background & Orientation • AI research on – belief representation & reasoning (R&R) and interpretation of propositional attitude reports (PA reports), with special attention to nested beliefs and default reasoning, … … and using a thorough mix of simulation theory and aspects of theory-theory – metaphor understanding (and soon, generation), … using a pretence-based / fictionalist approach • Developed running system, ATT-Meta, doing both belief R&R and metaphor understanding. • Attention to – metaphor embedded within belief (e.g., from metaphor use within PA report complements) – belief R&R embedded within metaphor (e.g., arising from personification metaphor). • Theoretical work on relationships between various figures of speech, notably metaphor and metonymy, but also metaphor/simile, irony/hyberbole, metaphor/hyperbole. – Metaphor, metonymy etc. as NOT being coherent scientific categories: based on underlying dimensions that are the real thing.