As part of our DRiPS project, we are holding a workshop on Wednesday 9th and Thursday 10th of October at Ågrenska Villan, Gothenburg.

Programme (with abstracts where available) as follows:

Wednesday 9th October

9.00-9.30 Coffee

9.30-9.45 Ellen Breitholtz & Chris Howes [slides] -- Opening remarks

9.45-12.00 Session 1 (Chair: Ellen Breitholtz)

• Mary Lavelle & Chris Howes [slides] -- Turn-taking and other interactional issues in patients with schizophrenia (9.45-10.30)

We report on some findings from the DRiPS project that show patients' difficulties in interactions (with healthy control participants who are unaware of the patient's diagnosis) are bound up with differences in the way that the turn-taking system operates in dialogues with patients with schizophrenia.

• Kristina Lundholm-Fors [slides] -- Linguistic and extra-linguistic parameters for early detection of cognitive impairment (10.30-11.15)

Dementia is a neurodegenerative disease that is becoming more prevalent as the mean age of the global population rises. To provide persons with dementia with the best care and enable the alleviation of symptoms through lifestyle changes and access to medication, it is important to identify individuals at risk for dementia at an early stage. Previous research has shown that subtle linguistic changes may be present even decades before a persons is diagnosed with dementia. In this project, which is in its fourth and final year, we have studied and analysed the linguistic abilities of persons with mild cognitive impairment in comparison with healthy controls. Our aim has been to find linguistic indicators that may useful in identifying individuals that are at risk for developing dementia later in life. I will talk about our results, the tasks that we have used, how we have applied language technology tools to analyse the data, and how machine learning has been used to investigate which features are most useful in differentiating between the groups.

• Michel Musiol -- Improving the investigation of conversational discontinuities using the support of eye-tracking methods (11.15-12.00)

We claim that schizophrenia disease can be described as discourse disorder. Meanwhile, it’s very hard to catch and describe obviously (at that time) all the discursive impairments and disruptions that characterize SCZ discourse. At that time formalisms elaborated in SLAM project allowed us to point two kinds of rules that are violated by SCZ speakers in discourse or dialog (« Right frontier breaking » and « Rise through the structure with Inconsistency »). These violations seem to be specific to schizophrenic discourse and, more precisely, to schizophrenic paranoïd specific discourse. But our formalisms actually explain only 5 % or 10 %) of all of de disruptions or incongruities that we may find in the course of SCZ discourse. We present an experimentation procedure based on a double eye-tracking system and on neuropsychological assessement to highlight neurocognitive and cognitive aspects of discursive competence that could be associated to incoherence and disruptions of verbal exchange. Our hypothesis is that inhibition of frequency of Saccadic Eye Mouvements and fixation or gaze direction of the psychologist interlocutor are concerned with visual attention and help to extract further relevant dialogical sequences to improve formalization of pathological discourse.

12.00-13.00 Lunch

13.00-14.30 Session 2 (Chair: Mary Lavelle)

• Manuel Rebuschi [slides] -- About context, fiction, and schizophrenia (13.00-13.45)

Research about conversations involving schizophrenic persons reveals the importance of underspecification. Conversation interpretation and context thus play a crucial role. In this talk, I will propose an analysis of schizophrenic discourse as relying on a new kind of pragmatic context, inspired by analyses of fictional discourse.

• Alva Stråge -- Do mentally disordered people lack rationality and free will? (13.45 - 14.30)

Assessment of people’s mental health plays an important role in our society. Not only because people with mental disorders may need treatment, but also because they may behave in ways that are hard to predict, and difficult to understand. One aspect of why it is important to assess whether a person suffers from a mental disorder is related to questions of responsibility. It is a common idea that if someone commits a criminal or immoral action under the influence of a serious mental disorder, she is not fully responsible for her action. This, in turn, is often thought to be related to her ability to act freely: if she acts under the influence of a mental disorder, she does not (or so it is argued) act out of her own free will

In the talk, I will discuss whether free will is constrained in people who suffers from mental disorders, and if it is, what kind of free will we then have in mind. The idea that people lack free will when they suffer from a mental disorder is often suggested to be supported by ‘disordered behavior’, as for example linguistic behavior: for example, if someone suffers from schizophrenia, she may have experiences that are taken to be out of touch with reality and therefore we may think that her choices are not sufficiently governed by rationality and that she is not able to respond to reasons in the way required for being said to choose freely when deciding how to act.

I will argue that even though behavioral evidence provides important information about how a person perceives the world from her subjective point of view, and thereby inform us of how we can foresee her behavior and interact with her on those premises, it provides us with less information about rationality and free will. As far as someone responds to reasons that she perceives as relevant, her behavior may be rational in relation to those reasons, even though we do not acknowledge the reasons she responds to as ‘real’ reasons. Moreover, I will argue that the fact that she responds to reasons we do not acknowledge as real does not necessarily make her less free than people who respond to reasons we do think are real.

I will make a distinction between two different notions of free will, and argue that the kind of free will that can be claimed to be constrained by a mental disorder is not of the sort that provides a relevant difference between a mentally disordered person on the one hand, and a healthy person on the other hand, when it comes to questions of responsibility and retributive punishment. This conclusion has the implication that the fact that a person suffers from a mental disorder may be of less relevance than we usually think when it comes to questions of whether someone, for example, deserves punishment for her actions.

14.30-15.00 Fika

15.00-17.00 Session 3 (Chair: Chris Howes)

• Maria Boritchev and Maxime Amblard [slides] -- Semantic processing of (real-life) questions and answers (15.00-15.45)

Semantic processing of discourse is a well-established field in computational linguistics. Starting from real-life dialogue data and especially non-standard phenomena it implies, we work towards producing a formal, logical, compositional and dynamic model of dialogue. We focus in particular on questions, answers and the way they can be articulated.

• Robin Cooper, Ellen Breitholtz & Bill Noble [slides] --  Combining TTR and game theory in dialogue modelling (15.45-17.00)

In this talk, we consider the notion of conversational games in dialogue and social meaning games, as used by Burnett (2018) to account for dialogue updates in situations where different dialogue strategies may have different social implications. Work in third wave sociolinguistics has defined the social meaning communicated by a particular linguistic choice in terms of the persona it is associated with (Eckert, 2012). We argue that social meaning games as described by Burnett can be integrated in a general semantic theory of dialogue such as KOS (Ginzburg, 2012). Finally, we present a probabilistic model that formalises the social meaning of a topos, or pattern of reasoning, and characterises the effect of social signals as a Bayesian update to the perceived persona of the speaker. Our formalisation of social meaning draws inspiration from both third-wave sociolinguistics, and recent work on the linguistic implications of the category adjustment effect (Regier and Xu, 2017).

17.00-19.00 Pub / free time

19.00 Dinner (Familjen)

Thursday 10th October

9.15-9.45 Coffee

9.45-12.00 Session 4 (Chair: Robin Cooper)

• Staffan Larsson [slides] -- Do dialogues have content yet? (9.45-10.30)

An implicit assumption in most research on dialogue semantics (such as SDRT) is that words do not change their meaning during a dialogue. Hence, it is assumed that the meaning of a dialogue can be captured by a (possibly existensially quantified) conjunction of statements in first order logic (or some modal logic). In this paper, the notion of “the content of a dialogue” is shown to be problematic in light of the phenomena of semantic coordination in dialogue, and the associated notion of semantic plasticity – the ability of meanings to change as a result of language use in interaction. Specifically, it appears that any notion of content in dialogue based on classical model-theoretical semantics will be insufficient for capturing semantic plasticity and coordination. An alternative formal semantics, type theory with records (TTR) is briefly introduced and is show to be better equipped to deal with semantic coordination and plasticity. However, it is also argued that any account of content in dialogue which takes semantic coordination seriously will also need to consider the problems it raises inference and truth, as well as for the notion of coherence in dialogue.

• Philippe de Groote -- From dynamic logic to dialogue modelling (10.30-11.15)

In this talk, I will discuss some of the problems to be solved in order to derive dialogue models from a continuation-based dynamic logic.

• Sylvain Pogodalla [slides] -- A (hybrid) logical approach to frame semantics (joint work with Laura Kallmeyer and Rainer Osswald) (11.15-12.00)

Frames emerged as a representation format of conceptual and lexical knowledge. They are commonly presented as semantic graphs with labelled nodes and edges. Structuring the knowledge as frames offers a fine-grained and systematic decomposition of meaning. However, it does not come with a natural way to integrate logical operators such as quantifiers. The approach we propose stems from the observation that modal logic is a powerful tool for describing relational structures, including frames. We use its hybrid logic extension in order to incorporate quantification and thereby allow for inference and reasoning.

12.00-12.30 Maxime Amblard [slides] -- wrap-up

12.30-13.30 Lunch

Conference webpage: https://clasp.gu.se/news-events/workshop-on-dialogue-and-perception-2018

Call for papers: https://easychair.org/cfp/DaP2018

Submissions: https://easychair.org/conferences/?conf=dap2018

Dates: June 14-15, 2018

Venue: Wallenberg Conference Centre, University of Gothenburg, Sweden

Organised by: CLASP

The study of dialogue investigates how natural language is used in interaction between interlocutors and how coordination and successful communication is achieved. Dialogue is multimodal, situated and embodied, with non-linguistic factors such as attention, eye gaze and gesture critical to understanding communication. However, studies on dialogue have often taken for granted that we align our perceptual representations, which are taken to be part of common ground (grounding in dialogue, Clark, 1996). They have also typically remained silent about how we integrate information from different sources and modalities and the different contribution of each of these. These assumptions are unsustainable when we consider interactions between agents with obviously different perceptual capabilities, as is the case in dialogues between humans and artificial agents, such as avatars or robots.

Contrarily, studies of perception have focussed on how an agent interacts with and interprets the information from their perceptual environment. There is significant research on how language is grounded in perception, how words are connected to perceptual representations and agent’s actions and therefore assigned meaning (grounding in action and perception, Harnad, 1990). In the last decade there has been impressive progress on integrated approaches to language, action, and perception, especially with the introduction of deep learning methods in the field of image descriptions that use end-to-end training from data. However, these have a limited integration to the dynamics of dialogue and often fail to take into account the incremental and context sensitive nature of language and the environment.

The aim of this workshop is to initiate a genuine dialogue between these related areas and to examine different approaches from computational, linguistic and psychological perspectives and how these can inform each other. It will feature invited talks by leading researchers in these areas, and high level contributed papers, presented as posters, selected through open competition and rigorous review.

We will be holding a fortnightly dialogue reading group in CLASP (in the seminar room at Dicksonsgatan 4), on Fridays from 10.15-12.00.

Full details, including links to papers, can be found here.