Hendrik Knoche
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Hendrik Knoche has a total of 125 co-authors. Here are the top co-authors by number of shared publications:
- Bastian Ilsø Hougaard
Department of Architecture, Design and Media Technology, Aalborg University, 9000 Aalborg, Denmark
8 shared publications - Robert Mikulik
St Anne’s University Hospital Brno, Czechia (R.M.).
8 shared publications - Ann Morrison
School of Future Environments, Auckland University of Technology, Auckland 1010, New Zealand
6 shared publications - Hamzah Ziadeh
Aalborg University, Aalborg, Denmark
6 shared publications - Mads Jochumsen
Department of Health Science and Technology, Aalborg University, 9260 Gistrup, Denmark
6 shared publications - Milo Marsfeldt Skovfoged
Department of Architecture, Design, and Media Technology, Aalborg University, Aalborg, Denmark
5 shared publications
Latest Publications
- Open Access
- 3 November 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Human-Computer Interaction
- Vol. 10 (7) , 1-21
Diagnosing Performance in Invasion-Based Esports: The Esports Performance Screening (EPS) Framework
Coaching in esports continues to become more common, yet conceptual tools for coaches to analyse performance breakdowns in esports remain limited. Existing approaches lack a way to distinguish between mental errors (i.e., mistakes relate to strategy and tactics), and physical slips (i.e., motor execution). This paper introduces the Esports Performance Screening (EPS) framework, that integrates several frameworks and models from sport and computing science which can be used to analyse mental and motor performance in esports. The EPS framework organises performance into five interconnected levels: strategy, tactics, tasks, actions and operations. Across these levels, teams pursue forms of superiority that influence transitions between stability and instability during invasion-based esports competition. The framework supports two complementary modes of screening use: diagnostic during reflective review and real-time screening during live play. Through illustrative esports scenarios, we demonstrate how instability can originate at different hierarchical layers and how misattribution of failure can obscure underlying causes. The purpose of the EPS framework is to help coaches systematically diagnose performance breakdowns, and act as heuristics to aid gameplay analysis during a match. The paper contributes a theoretically grounded, title-agnostic esports framework to support systematic performance analysis and pedagogical development in invasion-based esports.
- 3 November 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Human-Computer Interaction
- Vol. 10 (7) , 1-21
Open AccessDiagnosing Performance in Invasion-Based Esports: The Esports Performance Screening (EPS) Framework
Coaching in esports continues to become more common, yet conceptual tools for coaches to analyse performance breakdowns in esports remain limited. Existing approaches lack a way to distinguish between mental errors (i.e., mistakes relate to strategy and tactics), and physical slips (i.e., motor execution). This paper introduces the Esports Performance Screening (EPS) framework, that integrates several frameworks and models from sport and computing science which can be used to analyse mental and motor performance in esports. The EPS framework organises performance into five interconnected levels: strategy, tactics, tasks, actions and operations. Across these levels, teams pursue forms of superiority that influence transitions between stability and instability during invasion-based esports competition. The framework supports two complementary modes of screening use: diagnostic during reflective review and real-time screening during live play. Through illustrative esports scenarios, we demonstrate how instability can originate at different hierarchical layers and how misattribution of failure can obscure underlying causes. The purpose of the EPS framework is to help coaches systematically diagnose performance breakdowns, and act as heuristics to aid gameplay analysis during a match. The paper contributes a theoretically grounded, title-agnostic esports framework to support systematic performance analysis and pedagogical development in invasion-based esports.
- Open Access
- 2 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-16
Overengineered Underexplanations: Why What-If XAI Systems Fail to Assist Clinicians with Prognostic Tasks Due to Missing Strategic Conformance
- 2 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-16
Open AccessOverengineered Underexplanations: Why What-If XAI Systems Fail to Assist Clinicians with Prognostic Tasks Due to Missing Strategic Conformance
- Open Access
- 21 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-17
Reaching the last mile? How Design Decisions Shape Medical Device Qualification and Classification of Patient Support Chatbots In The EU
Few AI chatbots have transitioned the ’last mile’ from research prototypes to clinical practice, partly because mapping conversational design decisions to regulatory requirements is complicated. This paper presents a design-oriented regulatory analysis of a chatbot supporting stroke survivors. Using Design Space Analysis operationalised through the QOC (Questions, Options, Criteria) framework, HCI researchers and legal scholars assess how variations in three conversational functions — health Q&A, recovery monitoring, and motivating medical adherence — influence qualification and classification under the European Medical Device Regulation (MDR), General Data Protection Regulation (GDPR), and AI Act (AIA). Our findings show that seemingly minor design decisions — such as role, actionability, and proactivity — influence regulatory obligations. The paper contributes insights into ten conversational design considerations influencing regulatory interpretation and demonstrates how QOC can support integrating legal considerations into human-centred design workflows. Finally, we suggest compliance-aware pathways to market, while reflecting on whether all chatbots should pursue the ’last mile’.
- 21 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-17
Open AccessReaching the last mile? How Design Decisions Shape Medical Device Qualification and Classification of Patient Support Chatbots In The EU
Few AI chatbots have transitioned the ’last mile’ from research prototypes to clinical practice, partly because mapping conversational design decisions to regulatory requirements is complicated. This paper presents a design-oriented regulatory analysis of a chatbot supporting stroke survivors. Using Design Space Analysis operationalised through the QOC (Questions, Options, Criteria) framework, HCI researchers and legal scholars assess how variations in three conversational functions — health Q&A, recovery monitoring, and motivating medical adherence — influence qualification and classification under the European Medical Device Regulation (MDR), General Data Protection Regulation (GDPR), and AI Act (AIA). Our findings show that seemingly minor design decisions — such as role, actionability, and proactivity — influence regulatory obligations. The paper contributes insights into ten conversational design considerations influencing regulatory interpretation and demonstrates how QOC can support integrating legal considerations into human-centred design workflows. Finally, we suggest compliance-aware pathways to market, while reflecting on whether all chatbots should pursue the ’last mile’.
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-17
"Alone and Adrift in Analytics" - Insights from Long-term Involvements with stroke Clinicians when Using Care Quality Monitoring Systems
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-17
Open Access"Alone and Adrift in Analytics" - Insights from Long-term Involvements with stroke Clinicians when Using Care Quality Monitoring Systems
- Open Access
- 7 March 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Computing for Healthcare
Supporting Understanding: Comparing Conversational Interviewing in Chatbot vs Human-Administered Patient-Reported Outcomes for Stroke Survivors
Hospitals use patient-reported outcome measurements (PROMs) to monitor stroke survivors post-discharge. Unfortunately, PROM items are prone to misinterpretation. Human-administered PROMs mitigate comprehension issues through conversational interviewing (CI), in which facilitators provide real-time clarification. However, chatbots have not yet adopted CI techniques, raising questions about whether CI-enabled chatbots can effectively support PROM administration. We conducted a controlled experiment with 18 stroke survivors comparing two CI clarification techniques —1) additional examples and 2) reflective prompts —against 3) no clarification (control) administered by both a chatbot and a human facilitator. Our findings demonstrate the feasibility of CI-enabled chatbots, as participants’ responses remained consistent across facilitators and clarification types. Participants exhibited better PROM answer behaviour when using the chatbot, with fewer anecdotal digressions and greater adherence to validated response options. Notably, when receiving CI support, participants stopped asking the human facilitator for clarification. Preferences for clarification varied by question type: participants favoured examples for physical activity questions, while no clarification was preferred for mental health–related items.
- 7 March 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Computing for Healthcare
Open AccessSupporting Understanding: Comparing Conversational Interviewing in Chatbot vs Human-Administered Patient-Reported Outcomes for Stroke Survivors
Hospitals use patient-reported outcome measurements (PROMs) to monitor stroke survivors post-discharge. Unfortunately, PROM items are prone to misinterpretation. Human-administered PROMs mitigate comprehension issues through conversational interviewing (CI), in which facilitators provide real-time clarification. However, chatbots have not yet adopted CI techniques, raising questions about whether CI-enabled chatbots can effectively support PROM administration. We conducted a controlled experiment with 18 stroke survivors comparing two CI clarification techniques —1) additional examples and 2) reflective prompts —against 3) no clarification (control) administered by both a chatbot and a human facilitator. Our findings demonstrate the feasibility of CI-enabled chatbots, as participants’ responses remained consistent across facilitators and clarification types. Participants exhibited better PROM answer behaviour when using the chatbot, with fewer anecdotal digressions and greater adherence to validated response options. Notably, when receiving CI support, participants stopped asking the human facilitator for clarification. Preferences for clarification varied by question type: participants favoured examples for physical activity questions, while no clarification was preferred for mental health–related items.
- Open Access
- 1 March 2026
- Published by Oxford University Press (OUP) in European Stroke Journal
- Vol. 11 (4)
Information leaflets vs artificial intelligence: comparing perceptions of stroke survivors and professionals in a mixed-methods study
Introduction: Stroke survivors describe insufficient access to information post-discharge. We aimed to compare user perceptions of stroke information from third-sector stroke websites with those generated by a widely accessible artificial intelligence (AI) tool and to summarise the attributes of the preferred stroke information formats. Patients and methods: UK third-sector stroke websites were searched for materials relevant to 15 questions asked by stroke survivors. ChatGPT-4o was used to generate responses. Stroke professionals (clinicians, researchers), stroke survivors and caregivers reviewed third-sector and AI responses, guessing the source and identifying their preferred text. Participants rated responses on scales of empathy, trustworthiness, reliability, comprehensibility and usefulness and provided justification. Proportions of preference and correct guesses, as well as mean ratings, were compared between groups. Framework analysis was used to identify the attributes of response formats preferred by stroke survivors. Results: Relevant responses were found for 13 (87%) of 15 questions. Across groups, 60 participants with a mean age of 44 (SD = 14) and 57% females, correctly identified 184/300 (61%) of AI responses, and preferred AI in 123/300 (46%) of the cases. The groups differed in preference, with clinicians being least likely to choose AI (34%), followed by stroke survivors (49%) and researchers (54%). Stroke professionals viewed third-sector responses as more empathetic. The themes of content, structure and tone of responses were described with an emphasis on clarity, conciseness and an approachable tone. Discussion: Artificial intelligence-generated responses to stroke questions were rated positively by stroke survivors and researchers, whereas stroke clinicians were more sceptical. Conclusion: Widely accessible generic AI tools have the potential to complement existing stroke information resources.
- 1 March 2026
- Published by Oxford University Press (OUP) in European Stroke Journal
- Vol. 11 (4)
Open AccessInformation leaflets vs artificial intelligence: comparing perceptions of stroke survivors and professionals in a mixed-methods study
Introduction: Stroke survivors describe insufficient access to information post-discharge. We aimed to compare user perceptions of stroke information from third-sector stroke websites with those generated by a widely accessible artificial intelligence (AI) tool and to summarise the attributes of the preferred stroke information formats. Patients and methods: UK third-sector stroke websites were searched for materials relevant to 15 questions asked by stroke survivors. ChatGPT-4o was used to generate responses. Stroke professionals (clinicians, researchers), stroke survivors and caregivers reviewed third-sector and AI responses, guessing the source and identifying their preferred text. Participants rated responses on scales of empathy, trustworthiness, reliability, comprehensibility and usefulness and provided justification. Proportions of preference and correct guesses, as well as mean ratings, were compared between groups. Framework analysis was used to identify the attributes of response formats preferred by stroke survivors. Results: Relevant responses were found for 13 (87%) of 15 questions. Across groups, 60 participants with a mean age of 44 (SD = 14) and 57% females, correctly identified 184/300 (61%) of AI responses, and preferred AI in 123/300 (46%) of the cases. The groups differed in preference, with clinicians being least likely to choose AI (34%), followed by stroke survivors (49%) and researchers (54%). Stroke professionals viewed third-sector responses as more empathetic. The themes of content, structure and tone of responses were described with an emphasis on clarity, conciseness and an approachable tone. Discussion: Artificial intelligence-generated responses to stroke questions were rated positively by stroke survivors and researchers, whereas stroke clinicians were more sceptical. Conclusion: Widely accessible generic AI tools have the potential to complement existing stroke information resources.
- Open Access
Information Leaflets vs Artificial Intelligence: Comparing Perceptions of Stroke Survivors and Professionals in a Mixed-methods Study
Background: Stroke survivors often describe problems of insufficient access to information post-discharge. Traditional resources may not meet their information needs, but Artificial Intelligence (AI) could play a role. Aims: To compare user perceptions of stroke information from third sector stroke websites with that generated by AI and summarize the attributes of the preferred stroke information formats. Methods: UK third sector stroke websites were searched for materials relevant to 15 questions commonly asked by stroke survivors. ChatGPT(-4o) was used to generate responses to these questions. Stroke professionals (clinicians, researchers), stroke survivors, and their caregivers reviewed third sector and AI responses, indicating the source of the response and their preferred text. Participants also rated responses on scales of empathy, trustworthiness, reliability, comprehensibility and usefulness, and provided free text comments. Proportions of preference and correctly guessed responses, as well as mean ratings, were compared between the groups. Framework analysis was used to identify the attributes of response formats preferred by stroke survivors. Results: Relevant responses were found for 13 (87%) out of 15 questions. Across groups, 60 participants with mean age of 44 (SD=14) and 57% females, correctly identified 184/300(61%) of AI responses, and preferred AI response in 123/300(46%) of the cases. The groups differed in their preference with clinicians being least likely to choose AI (34%), followed by stroke survivors (49%) and researchers (54%). All groups viewed third sector responses as more empathetic in tone. The themes of content, structure, and tone of responses were described by stroke survivors with the emphasis on clarity, conciseness, and approachable tone. Conclusions: AI-generated responses to stroke questions were rated positively by stroke survivors and researchers, whereas stroke clinicians were more sceptical. Given that stroke information materials are intended for people with lived experience of stroke, their input should be prioritized to inform development of new resources.
Open AccessInformation Leaflets vs Artificial Intelligence: Comparing Perceptions of Stroke Survivors and Professionals in a Mixed-methods Study
Background: Stroke survivors often describe problems of insufficient access to information post-discharge. Traditional resources may not meet their information needs, but Artificial Intelligence (AI) could play a role. Aims: To compare user perceptions of stroke information from third sector stroke websites with that generated by AI and summarize the attributes of the preferred stroke information formats. Methods: UK third sector stroke websites were searched for materials relevant to 15 questions commonly asked by stroke survivors. ChatGPT(-4o) was used to generate responses to these questions. Stroke professionals (clinicians, researchers), stroke survivors, and their caregivers reviewed third sector and AI responses, indicating the source of the response and their preferred text. Participants also rated responses on scales of empathy, trustworthiness, reliability, comprehensibility and usefulness, and provided free text comments. Proportions of preference and correctly guessed responses, as well as mean ratings, were compared between the groups. Framework analysis was used to identify the attributes of response formats preferred by stroke survivors. Results: Relevant responses were found for 13 (87%) out of 15 questions. Across groups, 60 participants with mean age of 44 (SD=14) and 57% females, correctly identified 184/300(61%) of AI responses, and preferred AI response in 123/300(46%) of the cases. The groups differed in their preference with clinicians being least likely to choose AI (34%), followed by stroke survivors (49%) and researchers (54%). All groups viewed third sector responses as more empathetic in tone. The themes of content, structure, and tone of responses were described by stroke survivors with the emphasis on clarity, conciseness, and approachable tone. Conclusions: AI-generated responses to stroke questions were rated positively by stroke survivors and researchers, whereas stroke clinicians were more sceptical. Given that stroke information materials are intended for people with lived experience of stroke, their input should be prioritized to inform development of new resources.
- Open Access
- 9 January 2026
- Published by S. Karger AG in Cerebrovascular Diseases
- p. 1-13
Virtual Assistant as a Companion for Stroke Survivors post Discharge: Thematic Analysis Highlights Expectations of Human Performance Pose a Key Barrier to Engagement
Introduction: Health-related information and support is scarcely available to stroke survivors following discharge from the hospital, largely due to limited access to health professionals. Digital tools like virtual assistants (VAs) can address some of these needs on demand in a home environment, and there already is evidence of their effective deployment in health contexts. However, to maximise how individuals engage with such technological tools, it is integral that their design be grounded in the needs and associated expectations of the target user population. Methods: We conducted workshops where stroke survivors roleplayed as VAs and users and provided feedback on their expectations of a stroke-specific VA. In our one-to-one interviews with stroke survivors, we simulated VA interactions with two functions: administering PROMs and asking questions. Then, in testing sessions, stroke survivors, family, and caregivers completed PROMs both using the VA and with a human, before sharing their thoughts about the VA. Results: Our thematic analysis yielded four themes with semantic and latent codes. Aligned with past research, we observed an unmet need for on-demand information and support following discharge. Participants expected the VA to provide on-demand information not typically available through healthcare interactions. Participants perceived the VA as human or as a personable animate object, and these anthropomorphic expectations represented the most challenging-to-address barrier to engaging with the VA. Conclusion: These qualitative findings validate that common expectations of digital tools in general adult populations also apply to stroke survivors, highlighting the importance of participatory design. Accounting for ethical and risk considerations, we propose future directions for the development of safe and useful conversational agents that can serve as companions post discharge.
- 9 January 2026
- Published by S. Karger AG in Cerebrovascular Diseases
- p. 1-13
Open AccessVirtual Assistant as a Companion for Stroke Survivors post Discharge: Thematic Analysis Highlights Expectations of Human Performance Pose a Key Barrier to Engagement
Introduction: Health-related information and support is scarcely available to stroke survivors following discharge from the hospital, largely due to limited access to health professionals. Digital tools like virtual assistants (VAs) can address some of these needs on demand in a home environment, and there already is evidence of their effective deployment in health contexts. However, to maximise how individuals engage with such technological tools, it is integral that their design be grounded in the needs and associated expectations of the target user population. Methods: We conducted workshops where stroke survivors roleplayed as VAs and users and provided feedback on their expectations of a stroke-specific VA. In our one-to-one interviews with stroke survivors, we simulated VA interactions with two functions: administering PROMs and asking questions. Then, in testing sessions, stroke survivors, family, and caregivers completed PROMs both using the VA and with a human, before sharing their thoughts about the VA. Results: Our thematic analysis yielded four themes with semantic and latent codes. Aligned with past research, we observed an unmet need for on-demand information and support following discharge. Participants expected the VA to provide on-demand information not typically available through healthcare interactions. Participants perceived the VA as human or as a personable animate object, and these anthropomorphic expectations represented the most challenging-to-address barrier to engaging with the VA. Conclusion: These qualitative findings validate that common expectations of digital tools in general adult populations also apply to stroke survivors, highlighting the importance of participatory design. Accounting for ethical and risk considerations, we propose future directions for the development of safe and useful conversational agents that can serve as companions post discharge.
- Open Access
Survey for Categorising Explainable AI Studies Using Data Analysis Task Frameworks
Research into explainable artificial intelligence (XAI) for data analysis tasks suffer from a large number of contradictions and lack of concrete design recommendations stemming from gaps in understanding the tasks that require AI assistance. In this paper, we drew on multiple fields such as visual analytics, cognition, and dashboard design to propose a method for categorising and comparing XAI studies under three dimensions: what, why, and who. We identified the main problems as: inadequate descriptions of tasks, context-free studies, and insufficient testing with target users. We propose that studies should specifically report on their users' domain, AI, and data analysis expertise to illustrate the generalisability of their findings. We also propose study guidelines for designing and reporting XAI tasks to improve the XAI community's ability to parse the rapidly growing field. We hope that our contribution can help researchers and designers better identify which studies are most relevant to their work, what gaps exist in the research, and how to handle contradictory results regarding XAI design.
Open AccessSurvey for Categorising Explainable AI Studies Using Data Analysis Task Frameworks
Research into explainable artificial intelligence (XAI) for data analysis tasks suffer from a large number of contradictions and lack of concrete design recommendations stemming from gaps in understanding the tasks that require AI assistance. In this paper, we drew on multiple fields such as visual analytics, cognition, and dashboard design to propose a method for categorising and comparing XAI studies under three dimensions: what, why, and who. We identified the main problems as: inadequate descriptions of tasks, context-free studies, and insufficient testing with target users. We propose that studies should specifically report on their users' domain, AI, and data analysis expertise to illustrate the generalisability of their findings. We also propose study guidelines for designing and reporting XAI tasks to improve the XAI community's ability to parse the rapidly growing field. We hope that our contribution can help researchers and designers better identify which studies are most relevant to their work, what gaps exist in the research, and how to handle contradictory results regarding XAI design.
- Open Access
Experience of Virtual Help in a Simulated BCI Stroke Rehabilitation Serious Game and How to Measure It
Designers of digital rehabilitation experiences can accommodate error-prone input devices like brain–computer interfaces (BCIs) by incorporating virtual help mechanisms to adjust the difficulty, but it is unclear on what grounds users are willing to accept such help. To study users’ experience of virtual help mechanisms, we used three help mechanisms in a blink-controlled game simulating a BCI-based stroke rehabilitation exercise. A mixed-method, simulated BCI study was used to evaluate game help by 19 stroke patients who rated their frustration and perceived control when experiencing moderately high input recognition. None of the help mechanisms affected ratings of frustration, which were low throughout the study, but two mechanisms affected patients’ perceived control ratings positively and negatively. Patient ratings were best explained by the amount of positive feedback, including game help, which increased perceived control ratings by 8% and decreased frustration ratings by 3%. The qualitative analysis revealed appeal, interference, self-blame, and prominence as deciding experiential factors of help, but it was unclear how they affected frustration and perceived control ratings. Building upon the results, we redesigned and tested self-reported measures of help quantity, help appeal, irritation, and pacing with game-savvy adults in a follow-up study using the same game. Help quantity appeared larger when game help shielded players from negative feedback, but this did not necessarily appeal to them. Future studies should validate or control for the constructs of perceived help quantity and appeal.
Open AccessExperience of Virtual Help in a Simulated BCI Stroke Rehabilitation Serious Game and How to Measure It
Designers of digital rehabilitation experiences can accommodate error-prone input devices like brain–computer interfaces (BCIs) by incorporating virtual help mechanisms to adjust the difficulty, but it is unclear on what grounds users are willing to accept such help. To study users’ experience of virtual help mechanisms, we used three help mechanisms in a blink-controlled game simulating a BCI-based stroke rehabilitation exercise. A mixed-method, simulated BCI study was used to evaluate game help by 19 stroke patients who rated their frustration and perceived control when experiencing moderately high input recognition. None of the help mechanisms affected ratings of frustration, which were low throughout the study, but two mechanisms affected patients’ perceived control ratings positively and negatively. Patient ratings were best explained by the amount of positive feedback, including game help, which increased perceived control ratings by 8% and decreased frustration ratings by 3%. The qualitative analysis revealed appeal, interference, self-blame, and prominence as deciding experiential factors of help, but it was unclear how they affected frustration and perceived control ratings. Building upon the results, we redesigned and tested self-reported measures of help quantity, help appeal, irritation, and pacing with game-savvy adults in a follow-up study using the same game. Help quantity appeared larger when game help shielded players from negative feedback, but this did not necessarily appeal to them. Future studies should validate or control for the constructs of perceived help quantity and appeal.