Publikationer från Stockholms universitet
Not a member yet
31237 research outputs found
Sort by
Hur språkstil påverkar användares upplevelser och förtroende för chatbotar
No full textIntroduktion - Följande uppsats undersöker huruvida olika språkstilar i chatbotar påverkar upplevelser och förtroende hos användare. Chatbotar har blivit en naturlig del av många digitala tjänster men samtidigt saknas det kunskap om hur användare uppfattar olika språkstilar i chatbotar. Att förstå dessa skillnader är viktigt eftersom språkstil kan påverka hur användaren upplever chatbotars trovärdighet. Frågeställningar - Denna studie baseras på två frågeställningar: Hur skiljer sig användarnas upplevelser av textsvar mellan Claude, ChatGPT och Google Gemini? På vilket sätt påverkar svarens stil deras förtroende och upplevelse av informationen? Metod - Studien har en kvalitativ ansats där sex deltagare fick interagera med tre olika chatbotar: Claude, Google Gemini och ChatGPT. Dessa tre chatbotar valdes eftersom de skiljer sig tydligt i språkstil, vilket är relevant för studiens syfte att undersöka hur olika språkstilar påverkar användarnas upplevelser och förtroende för informationen. Efter interaktionen genomfördes semistrukturerade intervjuer för att samla in deltagarnas upplevelser. Det insamlade materialet analyserades med en induktiv tematisk analys för att identifiera återkommande mönster i hur språkstil påverkar upplevelser och förtroende. Resultat - Resultatet visar tydliga skillnader i hur deltagarna upplevde chatbotarnas språkstil, struktur och ton. Tre huvudteman identifierades genom en tematisk analys av intervjuerna. Dessa teman är språkstil och ton, struktur och uppbyggnad samt förtroende och trygghet. En mänsklig, empatisk och personlig ton hos chatbotarna upplevs i många situationer som trygg och tillitsfull. I andra situationer föredras mer opersonliga och faktabaserade svar. När chatbotarna använder en tydlig struktur som exempelvis punktlistor och rubriker ökar läsbarheten. Samtidigt upplevs längre sammanhängande svar ofta mer som ansträngande, även om längre svar uppskattas i vissa sammanhang. Resultaten visar att deltagarnas förtroende och upplevelser av chatbotarna skiljer sig markant beroende på vilken språkstil chatbotarna använder i kombination med vilken fråga och situation det gäller. Diskussion - Studiens resultat överensstämmer till stor del med tidigare forskning om hur språkstil påverkar användares upplevelser och förtroende för chatbotar samtidigt som denna bild nyanseras. En mänsklig och personlig stil kan uppfattas som trygg och stöttande i vissa situationer men i andra lägen kan den uppfattas som opassande eller överdriven. Vid faktabaserade eller medicinska frågor föredras en saklig och neutral ton samt att chatbotarna tydliggör sina begränsningar som AI-chatbotar. Vidare visar resultaten att chatbotarnas språkstil generellt sätt inte motsvarar tidigare forsknings beskrivningar vilket tyder på att språkstil och användares upplevelser av den är situationsbaserad och föränderlig över tid.The following thesis examines how different language styles used by generative chatbots affect user’s experiences and trust. As chatbots have become a central part of society, the need to understand how users perceive different language styles, expressions and how their trust is affected is becoming more important than ever before. This study addresses two research questions: How do users’ experiences of text responses differ between Claude, ChatGPT and Google Gemini? How do the chatbots’ language styles affect users’ trust and perception of information? This study has a qualitative research design where participants interacted with three selected chatbots, followed by semi-structured interviews in order to capture their experiences with the chatbots more deeply. The collected data were analyzed through an inductive thematic analysis to find and identify patterns and themes related to the chatbots language styles, structures and trust from the users. The results show clear differences in how the participants perceived the chatbot’s language styles, tones, and text structures. Three central themes were discovered: language style and tone, structural presentation and trust and safety. While an empathetic and personal text response was experienced as supportive and trustworthy in some contexts, a more neutral and factual response was preferred in others. A clear structural text presentation enhanced readability while longer, continuous texts were perceived as more challenging to read. Although the longer texts were preferred in some contexts. Overall, users' trust and experiences with the chatbots were shown to be highly dependent on the context and what language style, structure and tone the chatbots used along with the users preferences and needs. The discussion highlights that although the findings mainly correspond with previous research, they also provide a more nuanced understanding of chatbot language use. The findings showed that users’ perceptions are situational and dependent on context, with personal and empathetic language styles being appreciated in some scenarios but potentially inappropriate in others. The result findings show that the perceived trustworthiness of chatbots are closely linked to how well their language style matches the interaction context and user needs and preferences
Systemkontroll eller ett verktyg för enskilda? Svensk tillsynskultur möter GDPR
No full textI svensk förvaltningsrätt har tillsyn länge använts som ett verktyg för att säkerställa att den beslutade politiken får sitt avsedda genomslag i praktiken. Tillsyn kan rikta sig mot tillsynsobjekten i sig, dess generella förutsättningar, rutiner och processer, liksom mot hanteringen av enskilda fall i tillsynsobjektens verksamhet. Medan den första inriktningen primärt har till syfte att säkerställa att det rättsliga systemet fungerar som avsett, systemkontroll, kan den senare inriktningen även fungera som ett verktyg för enskilda att få genomslag för sina intressen. På senare är har EU-rätten fått en allt mer betydelsefull roll på tillsynsområdet, inte minst vad gäller införandet av sanktionsavgifter. Ett av de mer långtgående exemplen på EUrättsliga sanktionsavgifter finns i EU:s dataskyddslagstiftning. Såväl artikel 8.3 i EU:s stadga om de grundläggande rättigheterna (rättighetsstadgan) som artikel 16.2 fördraget om den Europeiska unionens funktionssätt (funktionsfördraget) anger att oberoende myndigheter ska kontrollera att EUrättens skydd för personuppgifter efterlevs. I EU:s dagaskyddsförordning, GDPR, preciseras vilka befogenheter de behöriga myndigheterna, dataskyddsmyndigheterna, ska ha och hur de ska samarbeta med varandra och den europeiska dataskyddsstyrelsen (EDPB) för att uppnå en effektiv tillsyn.5 Ett utmärkande drag i tillsynen av EU:s dataskyddsrätt är den enskildes ställning. Skyddet av personuppgifter är naturligen nära kopplat till den enskildes individuella sfär, framför allt den personliga integriteten.6 De regler som de oberoende myndigheterna ska kontrollera är således utformade som personanknutna rättigheter, där artikel 8.1 i rättighetsstadgan anger att ”var och en har rätt till skydd av de personuppgifter som rör honom eller henne.” Den övergripande frågan som analyserats i detta bidrag är hur GDPR har påverkat enskildas ställning i tillsynsärenden hos nationella dataskyddsmyndigheter, med svensk rätt i fokus
Cases to enhance reasoning in mathematics and education for sustainable development
No full textThe need to develop cross-curricular teaching in sustainable development is immense, and mathematics education has an important role. However, there are only a few suggestions on what such teaching can look like for younger children. This short communication presents a theoretical analysis of six cases used in a research project focusing on teaching children between 4 and 8 years old. The chosen concept was sharing, a key concept in sustainable development also relevant to mathematics education. The cases used in the project are analysed regarding afforded mathematical properties and ethical theories. Each case is then commented on using the combined results from the project. The cases illustrate a plausible content, and together with teachers’ didactical choices, they form an appropriate content for teacher education and professional development courses
Spänningen mellan demokrati och säkerhet : - vid bildande av Myndigheten för Psykologiskt försvar
No full textThe purpose of the essay is to analyse tensions between liberal democratic values and security and see how these tensions were handled in the process when the Swedish Psychological defence agency was created. Three research questions were formulated. These were: 1. What central themes are recurring in the consultation responses in the creation of Psychological defence agency? 2. Have the central themes been taken care of, and if they were, in what way? 3. How does the agency’s final regulation differ from the presented draft for consultation in the parts affected by identified themes? The research was done by examining the consultation responses in the referral process with qualitative method. Five themes were identified where tensions between liberal democratic values and security were present. These were 1) whether the agency should be established or not 2) remarks asking for a more precise definition of the concept of information influence 3) notes on poorly defined new task of reporting incidents regarding information influence for many of the government agencies 4) the new agency’s proposed task of helping media outlets in handling information influence and 5) whether the new agency should be able to direct signals intelligence. The analysis shows that the establishment of a new government agency leads to (re)negotiation of tasks between agencies, but also that new tasks are assigned to the new agency. In this process some of the questions that the agency is supposed to handle are securitized, they become part of the bureaucratic apparatus. The analysis also shows that not all suggestions became polity, for instance the new agency did not get the right to direct signals intelligence. This was after critique from consultation respondents. Other remarks were not taken care of, for instance the wish of precision of the concept of information influence was not in place when the agency started.
Generativ Artificiell Intelligens: “En mer kunnig annan”?
No full textAnvändningen av generativa AI-verktyg ökar snabbt inom högre utbildning, särskilt inom akademiskt skrivande, vilket väcker frågor om studenters självständighet och lärande. Denna kvalitativa studie undersöker hur universitetsstudenter upplever användningen av generativ AI (GAI) som stöd under skrivprocessen och i sitt lärande. Den teoretiska grunden bygger på Vygotskijs sociokulturella teori, särskilt begreppet den proximala utvecklingszonen (ZPD), som erbjuder ett perspektiv på hur GAI kan fungera som ett kognitivt och pedagogiskt verktyg. Studien bygger på semistrukturerade intervjuer med sex universitetsstudenter från olika utbildningsområden i Sverige. En tematisk analys användes för att identifiera mönster och teman i materialet. Resultaten visar att studenterna främst använder GAI i början och slutet av skrivprocessen för idégenerering, strukturering av text och grammatisk granskning. Även om AI ofta uppfattas som en ”hjälpsam coach” eller kompletterande handledare, värderar studenterna fortfarande mänsklig handledning för en djupare akademisk förståelse och återkoppling. Samtidigt uppmärksammas risker som felaktig information vid användning av GAI. Diskussionen belyser både möjligheter och utmaningar med att använda GAI i akademiskt skrivande. Även om GAI kan öka självförtroendet och minska kognitiv belastning, finns risker som överdrivet beroende, minskad motivation och reducerad reflektion. Resultaten tyder på att AI-verktyg fungerar bäst som kompletterande stöd inom studenternas ZPD och att framtida utveckling av GAI bör sträva efter mer kontextuella, ämnesspecifika och individualiserade funktioner som bättre överensstämmer med utbildningens mål.The use of generative AI tools is rapidly increasing in higher education, particularly in academic writing, which raises questions about students' independent thinking and learning. This qualitative study explores how students in higher education experience the use of generative AI (GAI) tools as support during the writing process and in their learning. The theoretical foundation is based on Vygotsky's sociocultural theory, specifically the concept of the Zone of Proximal Development (ZPD) which offers a perspective on how GAI might function as a cognitive and pedagogical tool. The study is based on semi-structured interviews with six university students from various academic fields in Sweden. Thematic analysis was used to identify patterns and themes in the data. The results show that students primarily use GAI in the early and final stages of the writing process for idea generation, text structuring and grammar correction. While AI often is looked at as a “helpful coach” or supplementary guide, students still value human supervision for a deeper academic understanding and feedback. However, there is always a risk of misinformation using GAI which the students acknowledge. The discussion highlights both the opportunities and challenges of using GAI in academic writing. Although GAI can promote confidence and reduce cognitive load, it can also cause risks such as overreliance, reduced motivation and decreased reflection. The findings suggest that AI tools are best used as complementary support within the students ZPD and that future development of GAI should aim for contextual, subject specific and personalized features that better align with educational goals
Designing to Care for Nature
No full textEngaging with the environment has been a significant topic of research within human-computer interaction, and it is particularly of interest to the field of sustainable HCI and the recent development of more-than-human design. Such investigations are gaining relevance as environmental disruptions continue to intensify and expand, bringing a sense of urgency to how we approach the more-than-human world around us. This has led to various engineered solutions that centre on mitigating impact, which have been crucial in addressing ecological challenges. However, despite such efforts, environmental disruptions persist and worsen. This thesis frames the reason for this not around inadequate engineered solutions, but rather around how they fail to address the root issue – a disconnection from and neglect of nature. We maintain a form of connection by consistently being informed by how environmental adversities affect us and our surrounding natural ecosystems. This increases our awareness and leads to a deep concern for ourselves and the state of the environment. However, this can also foster a sense of angst, desensitisation, and immobilisation, preventing us from taking action. This thesis argues for a need to translate this concern for nature into perspectives and practices that allow both humans and the environment to flourish: a move to caring. In this work, I explore how such a transition from concern to care in human-nature relations can be fostered through design. Positioned within the more-than-human turn in HCI, this exploration is based on four empirical studies, constructed through design-led methodologies, including research through design, autobiographical design, and a critical approach. These studies (1) outlined a more-than-human care framework to understand how technology shifts human-nature relations, (2) engaged with the multiple temporalities of nature through technology and an embodied practice, and critically analysed (3) media artworks to explore material deconstructions of posthuman interventions of time and (4) the meanings embedded within alternative time representations. Following these four investigations, this thesis pinpoints three dimensions to consider when designing to care for nature: enactments of care, expressions of time, and explorations of materiality. Each dimension has distinct aspects that can be combined and used to engage with fundamental challenges in caring for nature, which include the differing temporal dynamics between humans, technology, and the environment, as well as the dematerialisation of technology, which distances us from the physical realities of the world. This thesis reflects on the outlined dimensions by proposing design pathways to bring nature into view, bring nature closer, and bring people closer to nature. It also discusses the tensions within the dynamics of care, nature, and technology. This thesis contributes to HCI and design research situated from a more-than-human standpoint. Addressing a transition in human-nature relations from concern to care is both abstract and complex, highlighting a need for plurality in perspectives, methods, and practices. This thesis embraces such pluralities and supports a design space that cultivates interdependent and careful interactions with nature
Assessing Organizational AI Readiness in the Finnish Public Sector: A Case Study of the Helsinki District Court
No full textArtificial intelligence (AI) is rapidly changing and transforming both the private and public sector with increasing experimentation and implementation following the release of ChatGPT in 2023. Although Finland ranks among the leading countries in technological development and AI adoption, it remains unknown whether individual public organizations especially within the judicial sector are organizationally ready to adopt AI-based solutions. This study examines the organizational readiness of the Helsinki District Court, Finland's largest district court, through the Technology-Organization-Environment (TOE) framework's Organization dimension. Using a single-case study design, the research applies quantitative descriptive approach with qualitative elements. The data were collected via an online survey distributed to court judges and secretaries, consisting of eighteen 5-point Likert-scale statements and three open-ended questions. The quantitative data were analyzed using descriptive statistics while the qualitative responses were analyzed through thematic analysis. The findings indicate that the respondents possess strong digital skills and generally view AI in a positive light expecting it to support rather than replace their work. However, the organizational readiness is hindered by limited training, insufficient resources and limited communication regarding the goals and vision within the organization. Overall, the research conducted showed the Helsinki District Court demonstrating moderate readiness for AI adoption with cultural and skill-based strengths while facing structural and resource-related barriers
Synliga, men inte fullt erkända. Utsatta, men inte fullt skyddade. : En kritisk diskursanalys av remissyttranden till Ds 2022:18 Straffansvar för psykiskt våld
No full textUnder de senaste decennierna har den kriminalpolitiska debatten präglats av ett allt hårdare klimat där snabba lösningar premieras och signalerar handlingskraft. “Hårdare tag”-retorik har dominerat, där förslaget om sänkt straffbarhetsålder tydligt illustrerar det. I detta politiska landskap har psykiskt våld långsamt fått alltmer uppmärksamhet, vilket slutligen lett till ett förslag om att kriminalisera det. År 2022 skickades Ds 2022:18 Straffansvar för psykiskt våld ut på remiss och år 2025 skickades en reviderad version ut på remiss. Föreliggande studies syfte var att kritiskt analysera remissyttranden som inkommit till Ds 2022:18. Fokuset låg på hur psykiskt våld konstruerades av olika samhällsaktörer, vilka diskurser som framkom och hur detta kunde relateras till kvinnors prekaritet. För att kritiskt analysera språket användes den postmoderna feministiska kriminologin, där språk ses som ett centralt uttryck för makt. Detta kombinerades vidare med Judith Butlers begrepp precarity, som i korthet innebär att vissa liv betraktas som mindre skyddsvärda till följd av rådande samhällsnormer. Analysen visade att remissyttrandena präglades av tydliga diskursiva skiljelinjer. Kvinnoorganisationer fokuserade främst på hur en kriminalisering kunde samspela med befintliga könsmaktsordningar och därigenom påverka våldsutsatta kvinnors position inom rättssystemet. Aktörer inom rättskedjan betonade istället frågor om rättssäkerhet och gränsdragning. Det placerade det psykiska våldet inom ett ramverk av institutionell hanterbarhet snarare än strukturell maktutövning. Studien visar att psykiskt våld ges olika betydelser och avgränsningar. Kvinnors utsatthet görs därigenom synlig i remissmaterialet, men erkännandet är villkorat och varierar mellan diskurser. Det får konsekvenser för hur våldets allvar och relevans förstås inom den kriminalpolitiska ramen
Gender and Its Discontents : Categorization, Miscategorization and Misgendering of Trans and Non-binary People
No full textNon-binary people, who identify with gender categories other than women or men, are invisible, miscategorized, and misgendered, leading to their marginalization. Moreover, some non-binary people have specific and defined gender identities, and others are less specific and less defined. Gender-inclusive language reforms can expand the set of available categories and challenge people’s binary understanding of gender, but they also provoke backlash. The aim of this thesis is to investigate the possibility of language to make space for both types of non-binary identities, as well as to understand the dynamics of perceivers that produce backlash to non-binary people. Moreover, this thesis also aimed to test a recent psychological framework which predicted that people whose gender identity is more central to their sense of self should have a stronger negative reaction to non-binary people. Study I reviewed large national surveys from around the world to assess how the gender identities and sexual orientation of respondents are measured. Results show that most surveys used binary response formats or offer inconsistent non-binary options. Some used more inclusive response options, but these still drew boundaries and implicitly communicated which kinds of non-binary identities could and could not exist. Study II examined how response formats influence the categorization of gender-ambiguous faces. Participants categorized faces using different response options, including binary-only options and formats that included non-binary or “I don’t know” categories. I tested both whether response options could reduce participants’ tendency to think of gender as distinct categories (i.e., categorical perception) and increase their tendency to categorize others as non-binary. Response formats did not reduce participants’ tendency to treat gender categories as distinct (i.e., categorical perception), but formats that explicitly included non-binary options increased non-binary categorizations, especially of androgynous faces. Study III examined whether exposure to the Swedish gender-inclusive pronoun hen affects categorization of faces as non-binary. I also tested whether participants with stronger gender binary beliefs and whose gender identity was more central to their sense of themselves were more likely to categorize others as non-binary. Participants exposed to hen were more likely to categorize faces as non-binary than those exposed to other pronouns or no pronouns. Moreover, participants who believed that gender was binary were less likely to categorize others as non-binary. Participants’ gender identity centrality, however, was largely unrelated to their likelihood of categorizing others as non-binary. Study IV examined the characteristics of targets and perceivers that predict various outcomes related to rejection of non-binary people (specifically misgendering, categorization, and liking). That is, I examined which identities and which gender expressions are rejected. Additionally, I examined whether participants with strong gender binary beliefs and gender identity centrality were more likely to reject non-binary targets. Non-binary targets were more misgendered and miscategorized (but not disliked) than cisgender targets. Moreover, cisgender-prototypical non-binary targets were more often misgendered and miscategorized (but not disliked) than androgynous targets. Participants’ gender binary beliefs strongly predicted all three outcomes, but gender identity centrality only predicted misgendering
Undervisningspraktiker på Komvux : En kvalitativ fallstudie om lärares undervisningspraktiker på Komvux i relation till vuxna elevers språkliga utveckling i svenska som andraspråk
No full textThis study examines which teaching practices teachers of Swedish as a Second Language (SVA) in adult education (Komvux) use to support adult second language learners’ language development, as well as which methods students themselves perceive as most supportive. The study is grounded in a sociocultural perspective and is based on qualitative interviews with four SVA teachers, group-based student reflections, and observations of four lessons. The material was analyzed using content analysis. The results show that linguistic scaffolding through oral and written feedback is the most central teaching practice in SVA instruction. This is expressed through reformulations, explanations, and repetitions, which relate to Vygotsky’s theory of the Zone of Proximal Development and Gibbons’ description of scaffolding. The study also demonstrates that modeling, varied teaching methods, and a safe classroom climate are important for promoting adult L2 development. Students particularly highlight modeling, oral interaction, and feedback as the most effective ways to support their language development. However, peer feedback appears to be a challenging strategy, as it requires a high degree of structured training and strong language proficiency. There is a need for further research on feedback practices and didactic adaptations for heterogeneous student groups, as this study contributes to a deeper understanding of how teaching practices in adult education are shaped and experienced