SweCog 2022, Abstracts of Accepted Papers

D2-1-6: From rubber hands to virtual hands - A critical examination of the processes underlying bodily illusions

Andreas Kalckert
Bodily illusions such as the rubber hand illusion are well-known paradigms within experimental psychology and cognitive neuroscience. These paradigms have gained in popularity, with new variations of the illusions introduced almost every year. These new variants may include different sensory information (e.g., movements instead of tactile input) or other manipulations of the body (e.g., shape or look of the hand). Likewise, these illusions have been deployed in virtual reality which allows further manipulations not permitted in real settings. Most researchers draw equivalencies between these different variations, concluding for example that the illusion in virtual reality works in a similar manner to the real setting. In this talk I like to highlight two general problems with these interpretations: first, the assumptions of the underlying perceptual and cognitive processes generating the illusion experience and second, the way these illusions are quantified and results are interpreted. I like to point to certain caveats in bodily illusion paradigms this way. These issues may be important to consider in future applications of bodily illusions.

D2-2-2: Artefactual ethics as opportunity for rethinking “natural” ethics

Joel Parthemore
This paper argues that, within the ethics community, the wider philosophical establishment and society in general, people have been far too lax about what to accept as morally “right” behaviour – far too quick to let themselves and, all too often, each other off the hook. By drawing comparisons to artefactual behaviour and the objections people raise to calling that behaviour the morally acceptable behaviour of authentic moral agents, this paper lays out a framework by which human ethics and metaethics can more fruitfully be approached. An earlier paper of ours argued that, for an action to be morally right, one must have a convergence of the right motivations, the right means, and the right consequences. The underlying insight is that deontological, virtue-ethics-based, and consequentialist accounts all have their necessary role to play, but each tends to get too focused on itself and its merits to the loss of the bigger picture; while utilitarian accounts, as perhaps the most prominent division within consequentialism, face the further problem of failing to allow for those occasions where the needs of the few, or the one, outweigh the needs of the many, as Ursula K. LeGuin (1973) so devastatingly addressed. Although the requirement to align motivations, means, and consequences may seem impossibly onerous, it need not be, provided one is prepared to allow that moral behaviour is far more difficult to achieve, either for artefacts or human beings, than it might seem at first glance. Mistakes will be made, and perhaps it matters more to take responsibility for those mistakes than to assure oneself, despite reasonable argument to the contrary, that one has avoided them. It is time to hold artefactual and natural agent alike to a higher standard. [PDF]

D2-3-1: The leader learns it all? Using the “Kaptein Morf” tablet game to examine how different roles in joint problem solving affect learning

Andreas Falck and Janne von Koss Torkildsen
Kaptein Morf is a morphology-based vocabulary learning game for children aged 7-9, as well as a research tool (Torkildsen et al. 2021). Here, we describe the conceptual design of a multiplayer extension to the game, in which children can solve tasks while jointly attending each other’s solutions. The scientific aim of the multi-player game is to address how different roles within the joint attentional exchanges affect learning outcomes. For some tasks within the game session, children will be assigned to be the “leader”, i.e. initiating a response and selecting the final solution, and for other tasks to be the “follower”, i.e. having only an advisory role in the problem-solving. Learning in these two conditions will be compared to a baseline of solo play. The game allows tracking of learning on the level of single task items, enabling within-subject manipulation of the “leading” and “following” roles within the same game session. This makes the game engaging for the children, while still maintaining precise experimental control of the children’s turn-taking. The multiplayer extension is currently in development, and the present poster demonstrates how joint attention will be implemented in the game, with focus on the roles as “leader” and “follower”.

D2-3-2: What is Reason in the age of Artificial Intelligence and predictive processing?

Anders Persson
From philosophical and scientific accounts dating back as long as the Ancients Greeks, upon until contemporary days, there are on the face of it similar dyadic accounts for what Human Reasoning is, and is not. For Plato and Aristotle, Reason was an intellectual activity aimed for truth and knowledge, limited to Intellectuals and contrasted with Workers aiming to satisfy their Desire. Similar distinctions can be found with Kant and Hume, being in the Sensible or Intelligible worlds, aiming to satisfy Thinking, or Passions. Entering the 20th-centurary you have the distinction between Intuition and Reason, or some kind of Critical Thinking, and more recent accounts such as Implicit and Explicit knowledge and processes, or the infamous System 1 and System 2. With Artificial Intelligence on the agenda it begs the question, more than ever, what makes up a cognitive, intelligent, reflecting, thinking and reasoning agent, and it is not all that clear what previous accounts make out of a “Reasoning process”. Accepting a distinction between Intuition and Reason, there is an interesting question what predictive processing as an account of the brain adds to—with a generative working model simulated onto the world, and corrected by errors, it may seem to mostly be about Intuition. But coupled with, so called, “Offline” mental simulations, it might be an interesting account for Reason, that also might account for common critique of Human Reasoning abilities. In the poster presentation I will try to account for these theoretical considerations, as part of an ongoing development with my thesis.

D2-3-3: The Design of Intelligent Virtual Agents Using User-Centered Design Methods

Emma Mainza Chilufya
This paper outlines my PhD thesis project about the design of intelligent virtual agents (IVA). Ferbs [6] defines an IVA as “a physical or virtual entity that can act, perceive its environment (in a partial way) and communicate with others, is autonomous and has skills to achieve its goals and tendencies”. IVAs have potential applications in many shared spaces, such as first-line customer support, guiding in museums, receptions, etc. The design of IVAs is multidisciplinary and focuses on different user-centred aspects such as presence, emotion, appearance, behaviour and dialogue. Yet, design choices regarding these aspects are often based on the “introspective examination of personal preferences”Isbister et al.[7] rather than any accurate reflection of the design goals or the qualities valued by the users. [PDF]

D2-3-4: The role of prior experience in understanding speech: computational and experimental approaches to vowel perception

Anna Persson
One of the central challenges for speech perception is that talkers differ in their pronunciations. This results in between-talker differences in the mapping from the acoustic signal onto linguistic categories and meanings. Yet, listeners are remarkably adept at overcoming the initial difficulty in understanding new talkers. Despite substantial progress, the mechanisms that underlie these adaptive abilities remain unknown. I will present the initial steps of my research on this question for the perception of English vowels. I develop computational models based on phonetic databases, and test their predictions in web-based perception experiments to investigate whether listeners learn and store talker- and group-specific phonetic representations, and how the answer to this question might depend on pre-linguistic normalization procedures (like those commonly used in phonetic research.

D2-3-5: Vocal Characteristics predict Accuracy in Eyewitness Testimony

Philip Gustafsson
In two studies, we examined if correct and incorrect testimony statements were produced with vocally distinct characteristics. Participants watched a staged crime film and were interviewed as eyewitnesses. Witness responses were recorded and analyzed along 16 vocal dimensions. A mega-analysis of the two datasets showed four distinct vocal characteristics of accuracy; correct responses were uttered with a higher pitch, a "fuller voice", higher speech rate and shorter pauses. Taken together, this study advances previous knowledge by showing that accuracy is not only indicated by what we say, but also by how we say it.

D2-3-6: Do objective judges become emotional?                                

Alexander Tagesson
Affective processes are an integral part of much of juridical decision-making. Several researchers claim that affective processes, such as empathy and compassion, are parochial and biased, creating inconsistent decision-making. Consistency becomes especially important in juridical contexts, where inconsistent decisions can undermine the rule of law. With this background in mind, we tested how affective information, e.g. characterizing someone in a positive or negative way, and affective processes, e.g. how much compassion or empathy was felt with someone, affected Swedish district court judges’ decision-making during remand proceedings. The judges were asked to make several decisions connected to different applications for remand orders. Cases were presented as short vignettes, paired with a picture of the defendant and were designed to resemble real remand proceedings. Specific cases were matched on judicially relevant information, but, importantly, were mismatched on affective information. This design allowed us to examine how affective information affect juridical decision-making. We used self-reported empathy towards defendants and victims to predict judicially relevant decision outcomes.

D2-3-7: What do our eyes say about our estimation strategies?                

Maybí Morell Ruiz
Numerical estimation, measured with the Number Line Estimation Task (NLET), has been related to mathematical competence (Schneider, 2018) and numerical knowledge development. In this, eye-tracking has shown promising results in developmental studies of number sense (Schneider, 2008) and knowledge of numerical magnitude in children. Combining embedded eye-tracking technologies in laptops and tablets, preschoolers’ development of numerical estimation can be evaluated and integrated in early math educational software by means of pedagogically adaptive algorithms. In this pilot study with 10 PhD-students, performance (AEE, Absolute Estimated Error) and estimation strategies (eye fixation patterns) were evaluated using a laptop setup with eye-tracking and an on-screen implementation of NLET in a bounded and unbounded condition. Results on performance show that the unbounded condition (M=8.9; SD=5.92) has a lager AEE than the bounded condition (M=4.6; SD=2.43), with a significant medium effect size difference between conditions (t(26)=3.40, p<.002, Cohen's d=0.65). Results for the estimation strategies replicate previous findings, with eye-fixation patterns in the bounded condition describing a W shape and the unbounded a systematic downward trend. A next step is to embed this NLET eye-tracking methodology in a play-&-learn game for preschoolers.

D2-3-8: Human Interaction with Autonomous Drone Swarms: Design and control challenges

Oscar Bjurling
To date, Human-Swarm Interaction (HSI) research has largely focused on different problem areas in isolation, missing potential interaction effects between drone swarm architecture designs, control methods, and user interfaces that impacts system (and interaction) complexity. This highlights the pressing need for a holistic research approach. There is also a need for work-driven (complementing technology-driven) research to ensure the usability of swarm systems. Therefore, the current research project explores how swarm systems can assist in actual work environments (like forest firefighting or maritime search and rescue (SAR) operations), what capabilities they require, what challenges they pose to their operators, and how to design useable and efficient human-swarm interfaces for these work contexts. The conceptual work carried out in the current research project suggests that, in a real work context, a (de)centralized hybrid-control approach is required to strike a balance between swarm autonomy and resilient operation on the one hand, and operator control, situational awareness, and mental workload on the other. The system and user interface design must allow for the traversal between and within system strata, ranging from swarms and subswarms to individual drones and their sensors or equipment. For instance, this is important to a SAR swarm operator who must delegate tasks and supervise individual and groups of drones during a mission. Other current project results suggests that the swarm metaphor is, in fact, antithetical to the mission and user requirements presented above, and that choir or orchestra are perhaps better metaphors for generating useful designs.

D3-S1-1: What kind of memory is memory of fiction?                                                

Pierre Gander
Much of information people encounter in everyday life is not factual, such as from movies, novels and computer games. In recent years, there has been an increase in research on fiction, but memory of fiction and effects of fiction has been treated as isolated phenomena. There is a need for a theoretical account of how memory of fictional information is related to other types of memory and which mechanisms allow people to separate fact and fiction in memory. In this theoretical work, we propose an extension of Rubin's dimensional memory model to account for memories of fictional information of events, places, characters, and objects. Further, we offer a set of proposed mechanisms involving various degrees of complexity and levels of conscious processing, that mostly keep fact and fiction separated, but also allow learning and misinformation from fiction: content-based reasoning, source monitoring, and an associative link from the memory to the concept of fiction. In this way, we characterize the processing of fiction as a fundamental cognitive process that is innate, culturally universal, spontaneous, and independent of medium and modality and whether the information is mediated or directly experienced.

D3-S1-2: Cognitive bias in social services CPS case argumentation

Leonard Ngaosuvan
Child protective service (CPS) cases concern taking children into protective custody. Generally, social services investigate and present arguments for protective custody in court. The present study investigated a new type of cognitive bias in social services CPS case argumentation. The bias was first detected in an actual CPS investigation. The present study investigated the external validity of the bias. Participants (N = 133) completed an online within-subjects experiment where they rated the plausibility of two illogical arguments' (Simple vs. Complex), and six distractor items. The simple argument was as follows: "A is taller than B, hence C is taller than B". The complex argument was an abbreviated version of the actual CPS case where the parents appeared to provide inadequate attachment with the child. Broken down, the complex argument had the same isomorphic structure as the simple argument. The results showed that complex argument was considered implausible by 53%, and the simple 79%. The same pattern was found among participants with relevant academic training (N = 42); social worker, lawyer, psychologist, and students of said topics), 52% and 83% respectively. The results are discussed in terms of a new cognitive bias, and cognitive overload.

D3-S1-3: Sexual Economics in Swedish Dating: Pity Poor Men                

Linus Holmberg
Sexual exchange theory (SET) is a controversial theory describing heterosexual partner selection in terms of economic market factors. This paper explores SET empirically in Sweden, one of the most financially equal nations in the world. Experiment 1, a vignette study with four dating profiles, tested whether access to resources increases male attractiveness. Experiment 2, a vignette study measured how justifiable men's disappointment was, depending on financial courtship investments in a failed courtship attempt. The results of Experiment 1 indicated that, even in Sweden, men with limited resources are considered less attractive. Male financial resources are not seen as a bonus, but rather a prerequisite. In Experiment 2, participants felt that it was not justifiable to be disappointed for men who were 'cheap' in courtship. These results indicate that SET is a useful theory, even in a relatively gender-equal society.

D3-3-1: Human-Centered AI for Personalization in Experiential Learning Environments

Magnus Johnsson
Recent developments in AI, Virtual and Augmented Reality, Robotics etc. together with an improved understanding of human cognition provide opportunities for re-thinking the approach to education. Not only are there potential for improved efficiency and a better use of resources through automatization, but there are also opportunities for pedagogical improvements to fill in gaps in existing educational approaches. In addition to improving and optimizing learning environments, AI and other technologies can be used to help students become more creative and enthusiastic, and to help them gravitate towards a suitable role in life. Knowledge about human nature, cognition and learning (such as experiential learning, active learning theory, constructivism theory, discovery learning theory, situated learning theory, social learning theory), and principles such as universal design together with AI and other technologies makes it possible to create a physical-virtual experiential learning platform that is flexible, adaptive and personalizable to the needs of individual students through human-centered AI. We will present our approach to creating such a learning platform that help job-seeking students gain experiential learning through industry projects, internships and active engagement with small and medium-sized enterprises, while smoothly gravitating towards a suitable position and better self-understanding in a way that creates win-win situations for all stakeholders involved no matter what the outcome.

D3-3-2: Online filters and social trust: why we should still be concerned about Filter Bubbles

Andreas Falck
Eli Pariser's notion of a “Filter Bubble” describes the effect of social media filters tuned to predict what types of online contents social media users are likely to interact with, and subsequently presenting “more of the same” in order to maximise clicks. The Filter Bubble concept originally fuelled worries that users will find themselves in positive feedback loops, becoming exposed mostly to content that they already agree with, subsequently missing out on news and information that would contradict their pre-existing views. This initial worry has subsequently been challenged, by research showing that the views and sources that social media users are exposed to are actually quite diverse. Here, we argue that the original “Filter Bubble” theory, as well as subsequent criticisms, rest on a too simplified model of human belief formation, in which information content is over-emphasised at the expense of social dynamics. We argue that filter bubbles are still problematic, as they moderate peer feedback in a way that distorts how we evaluate information together with others. [PDF]

D3-3-3: The Crisis of Trust in AI and Autonomous Systems                                

Amandus Krantz
The future is robotic. Already we are seeing how society is changing with self-driving cars and robots at hospitals and schools. The huge potential of autonomous systems (AS) is highly discussed. What is not highly discussed, and rarely even acknowledged, is the key role trust plays in realizing these benefits. No common way of defining, testing, or measuring it exists. I argue that this lack of research on trust, both in general but also in human-AS relations, may at best result in sporadic progress and adoption of these systems and may at worst lead to public disillusionment and abandonment, delaying the potentials of AS. [PDF]

D3-3-4: The influence of contextual variability on learning novel words: Does the type of variability matter?

Raphaël Fargier
Adults predominantly learn new vocabulary from reading, and contextual variability benefits such learning. Contextual variability often refers to the number of unique documents a new word appears in, or to the number of different topics covered by the texts. Additionally, visual variability has been found to benefit learning of object words in children. In particular, variability in irrelevant object features (e.g. presenting chairs in different colors and materials) help children determine the core features of the object (e.g. that the core feature of a chair is its shape, not its color or material). In the present study, we examine what features of variability facilitate learning of novel object words from narrative contexts. We manipulated variability in non-definitional object features (e.g. color, size) and variability in situational contexts in which new objects are experienced (e.g. characters, location). In web-based experiments, participants encountered novel words in blocks of three short fictional narratives, and then provided a written definition of that word. Pilot data showed that lexical recognition performance was at ceiling at the immediate test, still high a week later (follow-up), and better in the condition with variability in non-definitional object features. Definition scores indicated better learning of the core semantic features in the condition with the highest degree of variability, i.e. variability in both non-definitional and situational features. Results suggest that situational variability may hinder lexical retention but may support better identification of core semantic features.