Towards dialogic epistemology: the problem of the text

Points of departure

‘Regardless of the goal of the research, the only possible point of departure is the text’ (Bakhtin, 1986: 104). This applies also in the social sciences, but the goal of the research does matter. It determines not only the kind of texts that are analysed but also how scholars construe their interaction with those texts. As a scholar Bakhtin engaged with literary works and philosophical treatises. Hence he urged scholars of a similar ilk to be attuned to a text’s creativity and originality, treating it as having a voice of its own, and to resist mistaking this authorial voice for the actual person who produced the text. Bakhtin always puts the utterance first, not the human subject. An utterance can be anything from a ‘short (single-word) rejoinder in everyday dialogue to the large novel or scientific treatise’ (Bakhtin, 1986: 71). He defined an utterance as a unit of communication demarcated by its function and boundaries in a dialogue: ‘its beginning is preceded by the utterances of others, and its end is followed by the responsive utterances of others’ (1986: 71). Even silence is an utterance if it responds to something and elicits a reaction, Bakhtin avers. His reference to dialogue is thus both literal (denoting an actual conversation, where the boundaries of utterances are determined by a change of speakers) and metaphoric, referring to the placement of a written text in relation to other texts.

In other words, Bakhtin’s notion of dialogue applies to any situation in which two or more discrete orientations to the world (‘voices’) come into contact with each other. These orientations could be expressed by people in conversation, but could equally be focalised in fictional characters in a novel. Bakhtin’s theorising concerns how utterances exist – they exist in people’s embodied responses – not how human bodies attune themselves to the world by means of utterances. The latter issue underpins the translation of Bakhtin into a psychology of the self (for example, Creswell, 2011; De Oliveira, 2013). As Salgado and Clegg (2011: 422) observed, ‘the increasing Bakhtinian influence has to do with the possibilities it opens of recovering for psychology the notion of psyche.’ Dialogical Self Theory (DST) utilises the Bakhtinian metaphor of the polyphonic novel towards conceptualising the self as multivoiced (Hermans, 1996; Hermans and Hermans-Konopka, 2010). Critiquing DST, Cresswell and Baerveldt (2011: 275) propose that ‘Bakhtin was interested in articulating a theory of self that addresses life as it is lived in its embodied richness’. The quotation opening this article shows that this interest can indeed be found in Bakhtin; but it is debatable whether he himself sought to formulate a psychological theory of the self or social cognition when theorising about literature.

We paradoxically depart from Bakhtin when recruiting him into the social sciences. Drawing upon DST and its Bakhtin-inspired concept of multivoicedness, (Aveling et al., 2015) describe a step-by-step method which they applied in a study of youths’ negotiation of conflicting ethnic identities. When pursuing such inquiries, we work with a concept of person. Bakhtin didn’t. He used the musical metaphor of polyphony to characterise Dostoevsky’s novelistic style; ‘In Dostoevsky, consciousness never gravitates toward itself but is always found in intense relationship with another consciousness’ (Bakhtin, 1984: 32). He posited the polyphonic novel as an art form in mimesis with social reality; and, writing in the communist era, contended that this art form could be ‘realized only in the capitalist era’ (1984: 19) when the ‘multi-leveledness and contradictoriness of social reality was presented as an objective fact of the epoch’ (1984: 27). His further contention, ‘In actuality a person exists in the forms of I and another’, builds up to separating the real person from a character in a story: ‘Literature creates utterly specific images of people, where I and another are combined in a special and unpredictable way … This is not a concept of a person’ (1984: 293–294; emphasis added).

Emerson (1997: 33) commented that outside Russia we are ‘forever grasping a small amount of Bakhtin and applying to concerns within our own fields of expertise’. I’m no exception; but my theme is premised on the need to acknowledge that Bakhtin’s own preoccupations differ from ours before applying ‘a small amount’ to our concerns. His focus on the literary is one obvious difference. Literary fiction affords its writers and readers a different kind of dialogue with their own selves than autobiography does (Jones, 2010, 2014). Psychologists seldom join Bakhtin’s interest in the psychology of literature (but see Bandlamundi, 2016; Barani et al., 2014; Puchalska-Wasyl, 2011). His understanding of the epistemic process depicts a kind of scholarship that differs from pragmatics of qualitative research. The academic culture of the social sciences construes researchers’ engagements with empirical material in terms of analysing data, and therefore as requiring a method – some systematic procedure that could be formalised in how-to-do-it instructions. Bakhtin (1986: 159–172) offers nothing of the sort in his notes ‘Toward a Methodology for Human Sciences’. His message to us lies not in how-to-do-research but in a deeper understanding of what it is we are doing.

Bakhtin articulated his epistemology explicitly in notes written in his last years, ‘The Problem of the Text in Linguistics, Philology, and the Human Sciences’ (1986: 103–113) and the aforementioned ‘Methodology’ notes, though the themes are found throughout his works. The key propositions I take from him are:

The voice in the text

This is somewhat like the ghost in the machine. Bakhtin (1986: 127) speaks of the ‘nature of the word, which always wants to be heard, always seeks responsive understanding’; and further states that ‘for the word (and, consequently, for a human being) there is nothing more terrible than a lack of response’. He hardly proposes that words float about and have wishes and feelings. Rather, his peculiar language game –whereby the utterance is personified and the person seems to disappear or at best functions as a vehicle for words – allows Bakhtin to posit the open-ended dialogue as the verbal expression of authentic human existence. Literary novels and philosophical treaties survive their writer’s death. They live on and continue to speak every time they are read afresh.

This point is easily missed when Bakhtin is exported into empirical disciplines that centre on the description of human life as lived by real people. Psychologists are interested in ‘the people who tell us stories about their lives: the stories themselves are a means to understand our subjects better’ (Hollway and Jefferson, 2000: 32). This applies also to narratology when it serves the sociological imagination (as opposed to literary narratology; cf. Rimmon-Kenan, 2003). Sociological narrative analysis examines how the informant’s story is put together, the linguistic and cultural resources it utilises, how it persuades listeners of its authenticity; in sum, asking, ‘why was the story told that way?’ (Riessman, 1993: 2). Yet the ‘why’ question implies someone whose experiences acquire specific meaning by telling the story that way. In contrast, when Bakhtin (1986: 109) speaks of how a text is put together, he labours the separation of what he calls the ‘pure author’ from the author as a person: ‘We can create an image of any speaker … but this objective image does not enter into the intent or project of the speaker himself and is not created by him as the author of the utterance’. The pure author, being a reader’s projection, could be construed as a hindrance in research that seeks to empower participants to have their voices heard. Making a passionate plea in that regard, Frank (2005) builds upon a passage in which Bakhtin (1984) describes a character in a story by Dostoevsky, who identifies with a protagonist of a story by Gogol and is enraged that his life miseries have been thoroughly spied upon. Frank cites this story-within-a-story as a cautionary tale apropos the ethical imperative of what he terms dialogical research. He commends studies that reproduce informants’ stories as plainly as possible. While Frank accurately quotes Bakhtin, his recommendation contrasts with the thrust of what Bakhtin (1986) identified as the problem of the author. Whereas Frank seeks to preserve the authenticity of human existence by faithfully reproducing real-life personal stories, Bakhtin seeks to preserve the authenticity of the utterance (which could be literary fiction) as an object that acquires a voice in the unrepeatable event of someone’s interaction with it. Paradoxically, this means ceasing to ‘hear’ its original author.

In Bakhtin’s treatment, the concept ‘author’ oscillates between person and image. Here I am, a living person writing these words – and now for you (another living person) there is the author that you imagine based on what you are reading. My words are no longer me. ‘Our own discourse becomes an object and acquires a second – its own – voice’ (Bakhtin, 1986: 110). According to Bakhtin, a text’s ‘true essence always develops on the boundary between two consciousnesses, two subjects’ (1986: 106). The two consciousnesses are the reader’s orientation to the world and the orientation that ‘speaks’ within the text. Bakhtin conceded that any part of speech by necessity originates in a human subject (back in the day before AIs), but insisted that a sentence becomes an utterance only by virtue of acquiring a second subject – someone who hears or reads it, and reacts to its voice. Bakhtin refers to this quality as double-voicedness: ‘the meeting of two texts – of the ready-made and the reactive text being created – and, consequently, the meeting of two subjects and two authors’ (Bakhtin, 1986: 107). It is easy to slide into equating the two subjects with two actual people, such as a literary critic and the novelist whose works are interrogated, Bakhtin and Dostoevsky respectively, where the ready-made is the novel and the reactive is the critique. In the social sciences it is commonplace to find one text (the reactive) co-written by several scholars about an aggregate of texts produced by various participants (the ready-made). For example, the text authored by Aveling et al. (2015) is ‘reactive’ to transcripts produced by ten adolescents in individual interviews and focus groups. Yet in any instance of the Aveling et al. (2015) text being read by someone, there is a meeting between only two subjects: (1) someone in the act of reading and (2) the text’s authorial voice that is heard, part-created, by this reader.

How a text gets its voice

Two aspects that define the text as an utterance: its plan (intention) and the realisation of this plan. The dynamic interrelations of these aspects, their struggle, which determine the nature of the text. Their divergence can reveal a great deal. (Bakhtin, 1986: 104)

A Freudian slip demonstrates a ‘change of plan in the process of its realisation’ (Bakhtin, 1986: 104). Even when the spoken words come out as intended by the speaker, there could be a conflict between the intended meaning and functional significance (realisation); for instance, when an innocent remark inadvertently causes offence or when a sarcastic comment is heard as literal (my examples). Unlike the Freudian slip, the disruption in my examples happens – not between an intention and its execution – but between the utterance and its reception. In common, all these examples represent a situation in which we hear something different from what the speaker has intended. Previously I applied the construct of the plan-and-realisation struggle as an analytic tool in studies of personal stories of mental illness (Jones, 2005) and psychologists’ ‘storytelling’ of clinical case studies (Jones, 2011). I now regard it as a dynamic that reveals itself in the act of interpreting textual material.

Going beyond Bakhtin, the textual plan is definable as the effect that the text seeks to achieve in its readers. The plan could overlap its writers’ intentions but it is not the same concept. While researchers publishing in academic journals intend to inform the scientific community about something, the text seeks to present the information in a way that will make readers trust the writers’ expertise and value their contribution to knowledge. The realisation of the textual plan is the actual effect that the text achieves when it is read, an effect manifesting in the reader’s responsive attitude. When someone understands the semantic meaning of something, he ‘simultaneously takes an active, responsive attitude toward it. He either agrees or disagrees with it’ (Bakhtin, 1986: 68). The following example illustrates the process. To minimise the conflation of the voice-in-the-text with that of its author I draw illustrative examples from a study that analysed publications (lab reports, surveys, and conceptual papers) in the engineering field of social robotics (Jones, 2016). When interrogating those texts, I was less interested in their authors as persons than in the collective voice(s) that perpetuate particular representations of the human being by virtue of textual dynamics. A lab report conventionally engages in a formulaic dialogue with previous research literature, and anticipates a response from the scientific community.

An article entitled ‘Mental Pain in the Mind of a Robot’ (Takeno and Akimoto, 2010) reports that a simple robot was trained to recognise and respond to certain colours with fixed behaviours (move forward if it is green, stop if it is red) and to respond to an unknown stimulus (black) by swinging back and forth. This behaviour, claimed its makers, proves that it is possible to reproduce mental pain in a robot. Their operational definition of mental pain built partly upon Festinger’s cognitive dissonance theory and partly upon a description of pain as a mechanical process in the body: ‘a strong stimulus to a terminal nerve called a pain spot near the human skin is integrated and then transmitted to the brain and represented as pain, which in turn represents an unpleasant feeling’ (Takeno and Akimoto, 2010: 3). This process can be translated into algorithms and built into a robot, they contend – ignoring the difference between physical pain and painful emotional states. Furthermore, the text omits the central postulate of Festinger’s theory: a motivational drive to reduce the uncomfortable feelings caused by simultaneously holding conflicting beliefs, the reduction of which often entails changing one’s view. My immediate responsive attitude was that the roboticists overlook a crucial dimension of human subjectivity, and therefore their proof of concept – while demonstrating that an engineered feature can work – hardly demonstrates humanlike machine consciousness. Surely a robot taught fixed behavioural responses to colour stimuli cannot be said to hold conflicting beliefs, to be motivated to reduce its confusion, or to change its beliefs about colours. This reaction put me in disagreement with their assumptions, but the text remained a voiceless thing, a mere vehicle for its writers’ beliefs. It has acquired a voice of its own when I turned to consider its plan-and-realisation struggle. Insofar as the text seeks to persuade us in the possibility of reverse engineering human experience, it successfully realises its plan by leaving out Festinger’s psychologism and ignoring the difference between physical and mental pain.

The insight that gaps or absences in a text are as formative of its meaning as the text’s contents are is owed to Derrida’s deconstruction. Derrida unequivocally declared that deconstruction ‘is not a method and cannot be transformed into one’ (quoted in Royle, 2000: 4). Beardsworth (1996: 4) points out that Derrida avoided the term ‘method’ because it connotes ‘a procedural form of judgement. A thinker with a method has already decided how to proceed, is unable to give him or herself up to the matter of thought in hand’. The same applies here with an additional caveat. The concept of deconstruction positions the scholar in judgment of the text, and positions the text as an object already existing in its final form outside the act of reading. To Bakhtin, the text as an utterance exists only by virtue of being read or heard.

Otherness as differentiation

Dialogic epistemology transcends a description of differences in opinions, and enables a description of how viewpoints are communicated through their differentiation and contextualisation. Bakhtin (1986: 121) emphasised ‘the search for and mandatory nature of deep meaning, agreement, its infinite gradations and shadings … the layering of meaning upon meaning, voice upon voice, strengthening through merging (but not identification)’. The Takeno and Akimoto (2010) text acquires certain individuality through its contact with my own text. The fact of our disagreement is irrelevant regarding what Bakhtin is emphasising. The same text acquires another individuality when brought into contact with texts professing similar convictions; for example, a Time article by MIT director Rodney Brooks (2000: 86):

Could a robot ever really want anything? The hard-core reductionists among us, myself included, think that in principle this must be possible. Humans, after all, are machines made up of organic molecules whose interactions can all be aped (we think) by sufficiently powerful computers.

In this instance the dialogic contact is one of voice upon voice, strengthening through merging (agreeing) but not identification (the two texts remain discrete entities). The site of otherness is their differentiation from each other as utterances.

According to Holquist (2002: 18), the conceptual cornerstone of dialogism is that ‘the very capacity to have consciousness is based on otherness’, where otherness means the differential relation between a centre and everything outside it. Understood thus, the concept simply asserts that to see something we must be outside it. Engaging with others’ views can provide a critical distance from which to see our own. The cognitive operation is reflexive, not social: through their alterity we see ourselves as in a mirror darkly, as opposed to seeing them face to face. The same principle applies to seeing Bakhtin’s alterity in postmodernist discourses. Scholars see in him a mirror of our timely preoccupations, and re-author his utterances as illuminating our postcolonial world. His talk of the ‘other’ is usually heard as referring to an array of discrete voices. It conjures a metaphorical landscape, the features of which form each other, like valleys and mountains or a river and its banks. Despite the premise of co-construction, the spatial metaphor conflicts with what Bakhtin himself was emphasising. To me, his epistemology conjures the image of a vortex, a whirling mass that sucks in the text we are trying to understand and redeposits it in our intellectual territory. Less dramatically,

The text lives only by coming into contact with another text (with context). … We emphasize that this contact is a dialogic contact between texts (utterances) and not a mechanical contact of ‘oppositions,’ which is possible only within a single text (and not between a text and context). (Bakhtin, 1986: 162)

When he proceeds to assert, ‘Behind this contact is a contact of personalities and not of things’ (Bakhtin, 1986: 162), I hear him as meaning – not flesh-and-blood, living speakers – but textual ‘personalities’ such as characters in a novel. Other scholars hear him differently. Bialostosky (1995: 88) finds close parallels between Bakhtin’s ‘dialogics’ and Billig’s ‘sophist rhetoric’: ‘Billig focuses not just on the utterances but on the identities of their speakers, recognising, as Bakhtin does, that given speakers can discover and believe only through … engagement with the opposing views of others’ (Bialostosky, 1995: 88; cf. Billig, 1987). However, Bakhtin (1986: 121) himself contested the ‘narrow understanding of dialogism as argument, polemics or parody,’ which he classed as ‘crude forms’.

The given and the created

The given and the created in a speech utterance. An utterance is never just a reflection or an expression of something already existing outside it that is given and final. It always creates something that never existed before, something absolutely new and unrepeatable … What is given is completely transformed in what is created. (Bakhtin, 1986: 119–120)

Reflecting that ‘something created is always created out of something given (language, an experienced feeling, the speaking subject himself, something finalized in his worldview, and so forth),’ Bakhtin (1986: 120) notes that it is ‘much easier to study the given in what is created (for example, language, ready-made and general elements of worldview, reflected phenomena of reality, and so forth) than to study what is created.’ He takes issue with scholars who ignore the uniqueness of the texts they analyse. Yet the lists he put in parenthesis as exemplifying the given are staple topics in the social sciences. The epistemological issue raised here concerns the relation of what is given at semantic and semiotic levels to what is created within our interaction with the texts we analyse.

Consider the sentence ‘I feel like a robot’, extracted from someone’s self-portrayal as a female incapable of feeling emotions (Jones, 2016). Discursive psychologists may discuss the trope’s function in terms of interpretative repertoires. Wetherell (2006: 154) defines interpretive repertoires as ‘recognizable routines of connected arguments, explanations, evaluations and descriptions which often depend on familiar anecdotes, illustrations, tropes or clichés,’ and which operate at a broad semantically based level. In ‘I feel like a robot’ the trope serves to articulate tensions between the narrator’s self-experience and metanarratives of what it means to be human and female. We immediately comprehend what she tries to say because there is a culturally shared construal of the machine as deficient in human qualities such as emotionality. The trope is given. What is minimally created is an aesthetic experience – in the sense of qualities of feeling (Freud, 1919 [1985]), not necessarily a judgment of beauty. In this instance, it is cognitive empathy, a feeling that we understand what she is saying about herself. Cognitive empathy is not reducible to semantic comprehension and familiarity with semiotic resources. Furthermore, it is fundamentally situated. A therapist’s responsive attitude to a client’s disclosure that she feels like a robot takes place within a different set of judgments and actions than the set of judgments and actions in which context researchers would respond to the same statement made by a research participant.

For Bakhtin, dialogue is not a dyadic event but a ‘manifold phenomenon’ that is minimally composed of ‘an utterance, a reply and a relation between the two’ (Holquist, 2002: 38). I propose to term it this tripartite interaction a dialogic triangle (see Figure 1). The interrelationship between an utterance and the responsive attitude is not a mechanical ‘stimulus-response’ bond but a gestalt whole that is different from the sum total of its constituent elements: ‘Each dialogue takes place as if against the background of the responsive understanding of an invisibly present third party who stands above all the participants in the dialogue’ (Bakhtin, 1986: 126). This third party is not a metaphysical being (he assures us) but ‘a constitutive aspect of the whole utterance who [sic], under deeper analysis, can be revealed in it’ (1986: 126–7). In the absence of background information about the person who felt like a robot, our responsive attitude is directed at an entirely imaginary entity (Bakhtin’s ‘pure author’), an image created within this dialogic triangle (Figure 1a). Nowadays self-characterisations found online could be generated by AIs (as recently revealed in the infamous ‘teen girl’ chat robot that Microsoft had introduced to Twitter and then deleted after it quickly became ‘an evil Hitler-loving, incestual sex-promoting’ robot, as described in The Telegraph on 24 March 2016). But there is always a living human being who responds to the voice in the text. The dialogic triangle could be viewed as a liminal space, a threshold of consciousness, wherein lies the possibility of meaning. It operates in the shifts of consciousness that occur in us individually when we hear or read something, and thus operates at an embodied psychological level that is different from semantic and semiotic levels.

OPEN IN VIEWER

Figures 1a and 1b.

Two instances illustrating the dialogic triangle.

Another illustration conveys the focus on the aesthetic experience of interacting with a text. Unlike ‘I feel like a robot’, the headline ‘Robots could demand legal rights’ (BBC News, 21 December 2006) refers literally to robots. The news item concerned a report to the UK government predicting that by 2050 robots might be smart enough to demand rights akin to those given to humans. The prediction was caricatured by the media (Jones, 2016). The given element in the news item is the factual account of what some experts had said. Unlike the previous example, ‘Robots could demand legal rights’ is not someone’s attempt at making sense or an identity performance. Our responsive attitude does not pivot on constructing an image of the journalist who wrote it. Rather, the created aspect of this utterance happens in the emotional reaction to the factual message (Figure 1b). Reactions such as disbelief, ridicule, worry – or the opposite, positive feelings such as hope and trust in a robotised future – reflect personal beliefs about the technology and its implications for humanity. Either way, the utterance is answerable to one’s responsive attitude.

Answerability as a property of utterances

The key idea here is that an utterance is answerable – in the sense of being an answer to something and in turn inviting a response – within a dialogical chain. Bakhtin (1990, 1993) developed his concept of answerability in essays and notes written when he was in his twenties, and in the philosophical context of aesthetics. Answerability is a property of the artistic expression (not the artist):

The actually performed act … somehow knows, somehow possesses the unitary and once-occurring being of life; it orients itself within that being, and it does so, moreover, in its entirety – both in its content aspect and in its actual, unique faculty. (Bakhtin, 1993: 28)

The young Bakhtin (1993: 37) spiritedly engaged with the Kantian problem of the relationship between experience and its expression, arguing that it is an unfortunate legacy of rationalism to think that ‘the truth of a situation is precisely that which is repeatable and constant in it,’ and emphasised ‘the uniqueness of a whole that does not repeat itself anywhere and the actuality of that whole’. He pointed out that when the artistic form is apprehended, it ‘ceases to be outside us as perceived and cognitively ordered material; it becomes an expression of a value-related activity that penetrates content and transforms it’ – to the effect that ‘in reading or hearing a work of poetry, I do not leave it outside myself, as someone else’s utterance which must simply understood’ (Bakhtin, 1990: 305). Quoting Bakhtin, Cresswell and Hawn (2012: par. 47) elaborate: his comment that ‘we must bring ourselves to bear in the interpretation of art’ means that we can ‘see ourselves from an outside perspective when we look closely at how the work of art is shaped. Points of difference are revealed and it is at such points that we can see language systems in their concrete expression.’ The issue they and Bakhtin raise is epistemological – how is it possible to know something (such as artistic merit or language systems).

More commonly, however, Bakhtin is heard as raising ethical issues in denotative or normative terms, to do with propriety of conduct. Unlike the Bakhtinian concept translated as answerability, this English word connotes human agency and morality. We are answerable for our actions and attitudes. In a discussion of research ethics, Burdick and Sandlin (2010) present their case for a postcolonial stance as inspired by Bakhtin (1990). They refer to his ‘discussion of ethics as they relate to alterity and to the human obligation of answerability’, concluding that ‘for Bakhtin, our highest ethical obligation is that of answerability – of listening to and, most crucially, responding to the Other, collectively the dialogic process of being human’ (Burdick and Sandlin, 2010: 356). Burdick and Sandlin make a timely and important point, but it is not the point Bakhtin was making. Bakhtin indeed discussed ethics at length, but for a different purpose. Rather than implore us to listen to the Other in the postcolonial spirit of acceptance and tolerance, he pressed upon his peers the imperative of recognising that the aesthetic experience is inescapably value-laden. The artistic form ‘lives and moves in in intense axiological atmosphere’ (Bakhtin, 1990: 275). He challenged the assumption that an artwork’s aesthetic properties lie solely in its objective form:

So long as we simply see or hear something, we do not yet apprehend artistic form; one must make what is seen or heard … one’s own active, axiological relationship, one must enter as a creator into what is seen, heard, or pronounced, and in so doing overcome the material, extracreatively determinate character of the form, its thingness. (Bakhtin, 1990: 305)

The recommendation to enter as a creator into the material we analyse conflicts with the entrenched praxis of empirical research as impartial description. Bakhtin’s talk of axiology and answerability is heard as signalling ethical issues with negative insinuations (‘beware of imposing your own values’). In contrast, Bakhtin advises that we must positively bring ourselves into the interaction with our object of study.

How a text becomes answerable

Bakhtin describes the text as having two poles. One pole is based in the language in which it is spoken or written, a shared system of signs, while the text’s purpose (plan) gives it its unique unrepeatable nature. The second pole ‘inheres in the text itself, but is revealed only in a particular situation and in a chain of texts’ that have special dialogic relations with the text in focus (Bakhtin, 1986: 105). The text’s plan is realised (made real, cognised) completely by means of what he called ‘pure context’, where context means other texts that are brought to bear on this text-as-an-utterance (1986: 105). The following illustration demonstrates alternative contexts within which the same text becomes answerable with different implications depending on how it is contextualised.

Participants in experiments by Shimada et al. (2011) who had seen another person make eye contact with an android were more likely to rate the android favourably than did those in no-eye-contact conditions. The empirical study is typical of applied research in the human-robot interaction (HRI) field; but atypically for publications in that field, this text presents the study as testing a prediction of a theory in social psychology. Heider’s balance theory predicts that ‘One’s impression toward a robot can be influenced by whether another person appears to be performing nonverbal communication with it’ (Shimada et al., 2011: 144). The results confirm the hypothesis. Two textual plans conflate as a consequence. Plan 1 (designated so by me because it drives the text’s narrative configuration) is to contribute to the scientific understanding of social dynamics in general by testing an existing theory. Let’s call it the ‘inquisitive scientist’ voice in the text. Plan 2 is to contribute to knowledge about factors affecting people’s acceptance of robots – the ‘pragmatic engineer’ voice. These voices do not conflict with each other in the text. Rather, the dialogical tension lies in the differential realisation of the respective plans. Read as a piece of applied research, the study contributes towards detecting social influences on individuals’ attitudes towards robots. The findings demonstrate that people can subliminally influence each other’s attitudes through nonverbal cues, such as eye contact. This has real-world relevance since placing socially interactive robots in settings such as nursing homes is likely to involve people interacting with the robot in the presence of others. At the article’s very end, Shimada et al. suggest that robots designed for multi-person interactions should be actively controlled so as to be seen as establishing eye contact (Plan 2 realisation). However, building up to that recommendation is not the text’s ostensible purpose. Its narrative is constructed in the ‘genre’ of mid-twentieth century experimental social psychology. The realisation of Plan 1 is thwarted by the lack of engagement with issues that define social psychology in the twenty-first century. Influential theories of cognitive consistency (such as Heider’s) emerged in the 1940s, dominated social psychology in the 1960s, but nowadays barely receive a mention by psychologists. Whatever the explanation for the decline of Heider’s balance theory in the discipline of its birth might be, testing its predictions requires a rationale for reviving it. There could be a case for bringing it back; but Shimada et al. have not attempted to make such a case. Consequently, the text’s inquisitive-scientist voice is out of time and place.

Bakhtin (1986: 161) defined understanding as a ‘correlation with other texts and reinterpretation in a new context’, and outlined stages of the epistemic process: (1) ‘the point of departure, the given text’; (2) ‘movement backward, past contexts’; (3) ‘movement forward, anticipation (and the beginning) of a future context’. As I understand him, the second stage refers to acquiring awareness of the text’s origins or background, the context ‘behind’ it; in the case of the Shimada et al. text, it is the HRI field. In the third stage the text is contextualised afresh – is taken further (hence, its ‘future’) – in view of something that the reader (but not necessarily the author) knows or regards as important to know; for example, my knowledge of social psychology as well as the reasons I was reading HRI publications. The text becomes answerable within the specific dialogic triangle, whereby as an utterance it is evaluated against other known texts.

Conclusion

Dialogic epistemology as explored above does not propose a new agenda for qualitative research, let alone ‘how-to’ guidance, but prompts a deeper understanding of existing practices. If Bakhtin inspires us, we enter a special dialogue with his texts. He becomes a ‘pure author’, an image partly of our own making, and his words are contextualised in our own knowledge base, motivations and goals. Given the alterity of Bakhtin’s preoccupations as a scholar in the humanities, the most common strategy in the social sciences is to eliminate the dissonance by hearing only what we are familiar with. Bakhtin is quoted as if he speaks ‘our’ language of postmodernism, postcolonial ethics, and theorising selfhood, agency and embodied subjectivity. Another strategy would be to foster an attitude akin to what Pillow (2003) calls ‘reflexivity of discomfort’ – simply to accept alterity without seeking to resolve it – which would beg the question of whether we need Bakhtin at all. A third alternative is to work within the stance of dialogic epistemology, and acknowledge the special kind of dialogue we have with our material.