Detta arbete har utförts inom ramarna för SIDUS-projektet AIR och fokuserar på interaktion mellan människa och autonoma och sociala robotar. Inom fältet för människa- robotinteraktion (MRI) ökar medvetenheten kring hur viktigt en positiv användarupplevelse (eng. user experience, UX) av dessa interaktioner är. När intresset för UX blir större ökar också behovet av att kunna arbeta med det på ett korrekt och lämpligt sätt. Idag finns det ett stort behov av metoder och tekniker för UX-arbete som är anpassade efter detta komplexa gränssnitt. Det övergripande syftet med detta arbete är därför att minska detta behov genom både en teoretisk litteraturstudie samt ett empiriskt arbete. I litteraturstudien kunde endast två ramverk ämnade för UX-utvärdering av MRI identifieras, varav ett av dem, USUS-ramverket, anses erbjuda en god grund för arbete med UX-utvärdering inom MRI. Fokus för det empiriska arbetet har sedan varit att förbättra och modifiera detta ramverk genom att integrera UX-mål som en del av det. UX-mål pekas ut som en central del för all sorts UX-arbete och är något som också kan optimera de utvärderingar som sker. Därför bör det också vara en del av det UX-arbete som sker inom MRI-fältet. Detta presenteras sedan i en ny version av USUS-ramverket, kallat USUS Goals. Baserat på dessa teoretiska och empiriska studier presenteras sedan riktlinjer för hur det fortsatta arbetet med UX- utvärdering inom MRI bör ske. Slutresultatet visar bland annat att utmaningarna med att integrera UX som del av MRI-fältet är större än vad som först antagits. Utmaningen ligger inte endast i att skapa användbara och anpassade UX-metoder, det är snarare ett ömsesidigt ansvar för båda domänerna att mötas för att tillsammans adressera dessa utmaningar. / Action and Intention Recognition in human interaction with autonomous systems (AIR)

Over the last few years, there has been a growing interest for social robots withhuman-like behavior and their application in healthcare and education. However,there are still plenty of issues that needs to be resolved. One of these challenges isto enable the social robots to fill its role effectively, by creating engagement. In this report the study, that was conducted at IBM Sweden, aims to understandhow IBM Watson can be utilized in the Pepper robot; to engage and supportchildren from the Ronald McDonald House in Uppsala. This is a place for childrenwith health problems and their families to live temporarily. Furthermore, supportivebehaviors are investigated since such behaviors are suggested to be important toincrease the engagement. An initial prototype that used Watson's natural languageprocessing and Pepper, was developed based on user requirements gatheredthrough interviews using a User Centered Design methodology. The prototype wasiteratively developed, and a final evaluation was conducted that examined both theperception of the robot as well as the engagement it created. The evaluation showed that the children wanted to interact with the robot again andhighlighted that they were highly engaged. They perceived the robot as a friend andthe supportive behaviors such as giving praise, responding quickly and maintainingeye contact were most important. The main support the children wanted were tohelp them feel less lonely and the conclusion of this study is that this is a suitablegoal for a robot system.

Master of Science / Department of Mechanical and Nuclear Engineering / Akira T. Tokuhiro / Robots will eventually become common everyday items. However before this becomes a reality, robots would need to learn be socially interactive. Since humans communicate much more information through expression than through actual spoken words, expression recognition is an important aspect in the development of social robots. Automatic recognition of emotional expressions has a number of potential applications other than just social robots. It can be used in systems that make sure the operator is alert at all times, or it can be used for psycho-analysis or cognitive studies. Emotional expressions are not always deliberate and can also occur without the person being aware of them. Recognizing these involuntary expressions provide an insight into the persons thought, state of mind and could be used as indicators for a hidden intent. In this research we developed an initial multi-modal emotion recognition system using cues from emotional expressions in face and voice. This is achieved by extracting features from each of the modalities using signal processing techniques, and then classifying these features with the help of artificial neural networks. The features extracted from the face are the eyes, eyebrows, mouth and nose; this is done using image processing techniques such as seeded region growing algorithm, particle swarm optimization and general properties of the feature being extracted. In contrast features of interest in speech are pitch, formant frequencies and mel spectrum along with some statistical properties such as mean and median and also the rate of change of these properties. These features are extracted using techniques such as Fourier transform and linear predictive coding. We have developed a toolbox that can read an audio and/or video file and perform emotion recognition on the face in the video and speech in the audio channel. The features extracted from the face and voices are independently classified into emotions using two separate feed forward type of artificial neural networks. This toolbox then presents the output of the artificial neural networks from one/both the modalities on a synchronized time scale. Some interesting results from this research is consistent misclassification of facial expressions between two databases, suggesting a cultural basis for this confusion. Addition of voice component has been shown to partially help in better classification.

Machines are designed to be infallible, but what happens if they are suddenly struck by chronic mental decline such as dementia? In this research, a social robot has been transformed into a mild-stage Alzheimer’s patient. The ultimate goal is to use it as a training tool for caregivers and medical students, as well as to raise general awareness for the disease. In particular, the study aimed to identify how to simulate Alzheimer’s with a social robot and what the effects are on its conversation partners. Thanks to its properties, the back-projected robotic head Furhat was the ideal candidate to adopt the role of Max. The sources of inspiration derived from interviews and observations. A Wizard of Oz setup enabled a conversation between the character and the user, who was given the task of asking about the robot’s life. To allow for in-between subject comparisons, the set of 20 participants was a mixture of medical and non- medical students, as well as people who knew someone with dementia closely and those who never met any. The experience was evaluated through pre- and post-interviews along with user observations. The results indicate that the patient simulation was convincing, leading the users to treat the machine as a human being and develop an emotional bond to it. They remained patient in spite of the robot’s symptoms, which affirms its potential for educational use. After all, this project aims to inspire researchers to find solutions in unconventional ways.

The rapid development of social robots, designed to interact with humans, has led to increased research on user acceptance and emotions in human-robot interaction. Social acceptance is an important area to investigate if the development of social robots is to be useful. Investigating how people feel about social robots is one tool to assess acceptance toward them, and research has shown that positive emotions could invoke higher acceptance. Possible factors that have been shown to affect peoples’ attitudes regarding social robots is (1) the human-likeness and appearance of the robot and (2) the application area of the robot. Therefore, this thesis research questions address the effect of human-likeness and application areas of social robots on people's emotions. The findings indicate that in the context of companionship, people have varying emotional responses based on the appearance of the social robot. Highly human-like robots evoke more positive emotions, while low human-likeness robots elicit more negative emotions. This suggests that individuals prefer human-like social robots in intimate interactions like companionship. The results also reveal an effect of application areas, where people respond more positively to highly human-like robots used for tasks like lecturing students or companionship for older adults. Regarding less human-like social robots, people tend to respond with greater positive emotions when used within commerce. This suggests that a simpler-looking robot with low human-likeness is more suitable for commercial applications. Negative emotions expressed in the healthcare condition may reflect mistrust in robots' abilities and the sensitivity of the healthcare area. Developers and designers should consider the emotional responses that might be evoked by the task or appearance of the social robot, to ensure successful integration into society.

In everyday communication, humans utilise eye gaze due to its importance as a communication tool. As technology evolves, social robots are expected to become more adopted in society and, since they interact with humans, they should similarly use eye gaze to elevate the level of the interaction and increase humans’ perception of them. Previous studies have shown that robots possessing human-like gaze behaviour increase the interactants’ task performance and their perception of the robot. However, social robots must also be able to behave and respond appropriately when humans act inappropriately, and failure in doing so may normalize bad behaviour even towards other humans. Additionally, with the recent progress of wearable eyetracking technology, there is interest to see how this technology can be used to help generate human gaze into a robot. This thesis work investigates how the eye gaze behaviour from a human being can be modeled into the robot Furhat to behave more human-like in a confrontational scenario. It further investigates how a robot possessing the developed human-like gaze model compares to a robot using a believable heuristic gaze model. We created a pipeline which concerned selecting scenarios, conducting roleplays between actors of these scenarios to collect gaze, extracting and processing that gaze data and extracting probability distributions that the human-like model would utilise. Our model used frequencies to determine where to gaze and head rotation, while gamma distributions were used to sample gaze length. We then executed an online video study with the two robot conditions where participants rated either robot by filling out a questionnaire. The results show that while no statistical difference could be found, the human-like condition scored higher on the measures anthropomorphism/human-likeness and composure, whereas the heuristic condition rated higher on expertise and extroversion. As such, the human-like model did not yield a significant benefit on robot perception to opt for it. Still, we suggest that the pipeline used in this thesis work may help HRI researchers to perform gaze studies and possibly build a foundation for further development. / I vardaglig kommunikation använder människor sig av blickar på grund av dess betydelse som kommunikationsverktyg. Då teknologi ständigt utvecklas förväntas det att sociala robotar kommer att bli mer involverade i samhället, och eftersom de integrerar med människor så bör de på samma sätt använda sig av blickar och ögonrörelser för att höja nivån på interaktionen och därmed förbättra människors uppfattning av dem. Tidigare studier har visat att robotar som använder sig av blickar likt människor kan förbättra deltagarnas utförande av uppgifter samt deras uppfattning av roboten. Sociala robotar måste dock även kunna agera och svara på ett lämpligt sätt när människor beter sig olämpligt, och gör dem inte det finns risken att det olämpliga beteendet normaliseras, även i interaktioner med andra människor. Med de senaste framstegen av portabla eye-tracking enheter finns det därför ett intresse att se hur denna teknologi kan användas för att generera människolikt blickbeteende som sedan används i en robot. Denna studie undersöker hur en människas sätt att blicka och titta kan modelleras i roboten Furhat för att bete sig mer människolikt i ett scenario där konfrontation behövs. Studien undersöker dessutom hur en robot som bär ett människolikt blickbeteende jämför sig med en robot som bär ett trovärdigt heuristiskt blickbeteende. Vi skapade en struktur som involverade att välja scenarion, utföra rollspel mellan skådespelare i dessa scenarier för att samla data om deras blickmönster, extrahera och bearbeta denna data, och extraherade sannolikhetsfördelningar som den människolika modellen skulle använda sig av. Vår modell använde sig av frekvenser för att besluta var roboten skulle blicka, medan gammafördelningar användes för att generera blickens längd. Vi utförde därefter en videostudie online med de två robotvarianterna, där deltagare bedömde någon av robotarna genom att svara på en enkät. Resultaten visar att ingen statistisk signifikant skillnad kunde påvisas. Trender visade dock att modellen med människolik blickbeteende bedömdes högre i mätningen av attributerna antropomorfism/mänsklighet och fattning, medan den heuristiska modellen bedömdes högre i expertis och utåtvändighet. Därav erhöll den människolika modellen ingen signifikant framgång för att föredra den. Vi föreslår ändå att strukturen som användes i studien kan hjälpa MRI forskare att utföra studier som involverar blickbeteende hos människor, och möjligtvis bygga en grund för vidareutveckling av strukturen.

Achieving appropriate and natural backchanneling behavior in social robots remains a challenge in Human-Robot Interaction (HRI). This thesis addresses this issue by utilizing methods from Safe Reinforcement Learning in particular shielding to improve social robot backchanneling behavior. The aim of the study is to develop and implement a safety shield that guarantees appropriate backchanneling. In order to achieve that, a Recurrent Neural Network (RNN) is trained on a human-human conversational dataset. Two agents are built; one uses a random algorithm to backchannel and another uses shields on top of its algorithm. The two agents are tested using a recorded human audio, and later evaluated in a between-subject user study with 41 participants. The results did not show any statistical significance between the two conditions, for the chosen significance level of α < 0.05. However, we observe that the agent with shield had a better listening behavior, more appropriate backchanneling behavior and missed less backchanneling opportunities than the agent without shields. This could indicate that shields have a positive impact on the robot’s behavior. We discuss potential explanations for why we did not obtain statistical significance and shed light on the potential for further exploration. / Att uppnå lämpligt och naturligt upbbackningsbeteende i sociala robotar är fortfarande en utmaning i Människa-Robot Interaktion (MRI). Den här avhandlingen tar upp detta problem genom att använda metoder från säker förstärkningsinlärning i synnerhet avskärmning för att förbättra sociala robotars upbbackningsbeteende. Syftet med studien är att utveckla och implementera en säkerhetsavskärmning som garanterar lämplig upbbackning. För att uppnå det, tränas ett återkommande neuralt nätverk på en människa-människa konversationsdatamängd. Två agenter byggs; en använder en slumpmässig algoritm för att upbbacka och en annan använder avskärmninng ovanpå sin algoritm. De två agenterna testas med hjälp av ett inspelat mänskligt ljud och utvärderas senare i en användarstudie med 41 deltagare. Resultaten visade inte någon statistisk signifikans mellan de två skicken, för den valda signifikansnivån < 0, 05. Vi observerar dock att agenten med avskärmning hade ett bättre lyssningsbeteende, mer lämplig upbbackningsbeteende och missade mindre upbbacknings-möjligheter än agenten utan avskärmning. Detta kan indikera att avskärmning har en positiv inverkan på robotarnas beteende. Vi diskuterar potentiella förklaringar till varför vi inte fick statistisk signifikans och belyser potentialen för ytterligare utforskning.

This study investigates if preferences for mental capacities in robots vary across different application domains and identifies influential factors, both in individuals’ characteristics and attributes specific to each robot domain. Employing a between-subject design, participants (N=271) completed a survey collecting both quantitative and qualitative data on preferences for 12 mental capacities across six robot types situated in a specific application domain (medicine, defense, household, social, education, customer service). Half of the mental capacities align with each dimension (experience and agency) in the two-dimensional model of mind (Gray et al., 2007; McMurtrie, 2023). Key findings reveal a general preference for high agency ability and low experience ability across all application domains. Exceptions were found in preference for lower agency ability in the cleaning robot and higher experience ability in the companion robot. Qualitative analysis indicates a desire for objective and logical robots functioning without emotions, while demonstrating empathy for human emotions. Additionally, gender and educational background emerged as factors influencing preference for lower experience abilities in robots. While previous research has mainly focused on attribution of mental capacities to technical agents, this study provides insights into human preferences and factors affecting them. These insights can guide responsible and ethics-driven development and design of robot technology within the field of human-robot interaction.

Robots and virtual agents grow rapidly in behavioural sophistication and complexity. They become better learners and teachers, cooperators and communicators, workers and companions. These artefacts – whose behaviours are not always readily understood by human intuition nor comprehensibly explained in terms of mechanism – will have to interact socially. Moving beyond artificial rational systems to artificial social systems means having to engage with fundamental questions about agenthood, sociality, intelligence, and the relationship between mind and body. It also means having to revise our theories about these things in the course of continuously assessing the social sufficiency of existing artificial social agents. The present thesis presents an empirical study investigating the social influence of physical versus virtual embodiment on people's decisions in the context of a bargaining task. The results indicate that agent embodiment did not affect the social influence of the agent or the extent to which it was perceived as a social actor. However, participants' perception of the agent as a social actor did influence their decisions. This suggests that experimental results from studies comparing different robot embodiments should not be over-generalised beyond the particular task domain in which the studied interactions took place.

The topic of this thesis is AI recruitment chatbots, digital discrimination, and data feminism (D´Ignazio and F.Klein 2020), where I aim to critically analyze issues of bias in these types of human-machine interaction technologies. Coming from a professional background of theatre, performance art, and drama, I am curious to analyze how using AI and social robots as hiring tools entails a new type of “stage” (actor’s space), with a special emphasis on social acting. Humans are now required to adjust their performance and facial expressions in the search for, and approval of, a new job. I will use my “theatrical glasses” with an intersectional lens, and through a methodology of cultural analysis, reflect on various examples of conversational AI used in recruitment processes. The silver bullet syndrome is a term that points to a tendency to believe in a miraculous new technological tool that will “magically” solve human-related problems in a company or an organization. The captivating marketing message of the Swedish recruitment conversational AI tool – Tengai Unbiased – is the promise of a scientifically proven objective hiring tool, to solve the diversity problem for company management. But is it really free from bias? According to Karen Barad, agency is not an attribute, but the ongoing reconfiguration of the world influenced by what she terms intra-actions, a mutual constitution of entanglement between human and non-human agencies (2003:818). However, tech developers often disregard their entanglement of human-to-machine interferences which unfortunately generates unconscious bias. The thesis raises ethical questions of how algorithmic measurement of social competence risks holding unconscious biases, benefiting those already privileged or those acting within a normative spectrum.