\n \n \n
\n
\n\n \n \n \n \n \n \n Role of Reasoning in LLM Enjoyment Detection: Evaluation Across Conversational Levels for Human-Robot Interaction.\n \n \n \n \n\n\n \n Marcinek, L.; Irfan, B.; Skantze, G.; Pereira, A.; and Gustafsson, J.\n\n\n \n\n\n\n In Béchet, F.; Lefèvre, F.; Asher, N.; Kim, S.; and Merlin, T., editor(s),
Proceedings of the 26th Annual Meeting of the Special Interest Group on Discourse and Dialogue, pages 573–590, Avignon, France, 2025. Association for Computational Linguistics\n
\n\n
\n\n
\n\n
\n\n \n \n
Paper\n \n \n\n \n\n \n link\n \n \n\n bibtex\n \n\n \n \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@inproceedings{marcinek_role_2025,\n\taddress = {Avignon, France},\n\ttitle = {Role of {Reasoning} in {LLM} {Enjoyment} {Detection}: {Evaluation} {Across} {Conversational} {Levels} for {Human}-{Robot} {Interaction}},\n\tshorttitle = {Role of {Reasoning} in {LLM} {Enjoyment} {Detection}},\n\turl = {https://aclanthology.org/2025.sigdial-1.46/},\n\tabstract = {User enjoyment is central to developing conversational AI systems that can recover from failures and maintain interest over time. However, existing approaches often struggle to detect subtle cues that reflect user experience. Large Language Models (LLMs) with reasoning capabilities have outperformed standard models on various other tasks, suggesting potential benefits for enjoyment detection. This study investigates whether models with reasoning capabilities outperform standard models when assessing enjoyment in a human-robot dialogue corpus at both turn and interaction levels. Results indicate that reasoning capabilities have complex, model-dependent effects rather than universal benefits. While performance was nearly identical at the interaction level (0.44 vs 0.43), reasoning models substantially outperformed at the turn level (0.42 vs 0.36). Notably, LLMs correlated better with users' self-reported enjoyment metrics than human annotators, despite achieving lower accuracy against human consensus ratings. Analysis revealed distinctive error patterns: non-reasoning models showed bias toward positive ratings at the turn level, while both model types exhibited central tendency bias at the interaction level. These findings suggest that reasoning should be applied selectively based on model architecture and assessment context, with assessment granularity significantly influencing relative effectiveness.},\n\turldate = {2026-07-25},\n\tbooktitle = {Proceedings of the 26th {Annual} {Meeting} of the {Special} {Interest} {Group} on {Discourse} and {Dialogue}},\n\tpublisher = {Association for Computational Linguistics},\n\tauthor = {Marcinek, Lubos and Irfan, Bahar and Skantze, Gabriel and Pereira, Andre and Gustafsson, Joakim},\n\teditor = {Béchet, Frédéric and Lefèvre, Fabrice and Asher, Nicholas and Kim, Seokhwan and Merlin, Teva},\n\tyear = {2025},\n\tpages = {573--590},\n}\n\n\n\n\n
\n\n\n
\n User enjoyment is central to developing conversational AI systems that can recover from failures and maintain interest over time. However, existing approaches often struggle to detect subtle cues that reflect user experience. Large Language Models (LLMs) with reasoning capabilities have outperformed standard models on various other tasks, suggesting potential benefits for enjoyment detection. This study investigates whether models with reasoning capabilities outperform standard models when assessing enjoyment in a human-robot dialogue corpus at both turn and interaction levels. Results indicate that reasoning capabilities have complex, model-dependent effects rather than universal benefits. While performance was nearly identical at the interaction level (0.44 vs 0.43), reasoning models substantially outperformed at the turn level (0.42 vs 0.36). Notably, LLMs correlated better with users' self-reported enjoyment metrics than human annotators, despite achieving lower accuracy against human consensus ratings. Analysis revealed distinctive error patterns: non-reasoning models showed bias toward positive ratings at the turn level, while both model types exhibited central tendency bias at the interaction level. These findings suggest that reasoning should be applied selectively based on model architecture and assessment context, with assessment granularity significantly influencing relative effectiveness.\n
\n\n\n
\n\n\n
\n
\n\n \n \n \n \n \n \n Speech-to-Joy: Self-Supervised Features for Enjoyment Prediction in Human–Robot Conversation.\n \n \n \n \n\n\n \n Santana, R.; Irfan, B.; Lagerstedt, E.; Skantze, G.; and Pereira, A.\n\n\n \n\n\n\n In
Proceedings of the 27th International Conference on Multimodal Interaction, pages 238–248, Canberra Australia, 2025. ACM\n
\n\n
\n\n
\n\n
\n\n \n \n
Paper\n \n \n\n \n \n doi\n \n \n\n \n link\n \n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@inproceedings{santana_speech--joy:_2025,\n\taddress = {Canberra Australia},\n\ttitle = {Speech-to-{Joy}: {Self}-{Supervised} {Features} for {Enjoyment} {Prediction} in {Human}–{Robot} {Conversation}},\n\tisbn = {979-8-4007-1499-3},\n\tshorttitle = {Speech-to-{Joy}},\n\turl = {https://dl.acm.org/doi/10.1145/3716553.3750747},\n\tdoi = {10.1145/3716553.3750747},\n\tlanguage = {en},\n\turldate = {2026-07-25},\n\tbooktitle = {Proceedings of the 27th {International} {Conference} on {Multimodal} {Interaction}},\n\tpublisher = {ACM},\n\tauthor = {Santana, Ricardo and Irfan, Bahar and Lagerstedt, Erik and Skantze, Gabriel and Pereira, Andre},\n\tyear = {2025},\n\tpages = {238--248},\n}\n\n\n\n\n
\n\n\n\n
\n\n\n
\n
\n\n \n \n \n \n \n \n Online Prediction of User Enjoyment in Human-Robot Dialogue with LLMs.\n \n \n \n \n\n\n \n Janssens, R.; Pereira, A.; Skantze, G.; Irfan, B.; and Belpaeme, T.\n\n\n \n\n\n\n In
Proceedings of the 2025 ACM/IEEE International Conference on Human-Robot Interaction, of
HRI '25, pages 1363–1367, Melbourne, Australia, 2025. IEEE Press\n
\n\n
\n\n
\n\n
\n\n \n \n
Paper\n \n \n\n \n\n \n link\n \n \n\n bibtex\n \n\n \n\n \n \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@inproceedings{janssens_online_2025,\n\taddress = {Melbourne, Australia},\n\tseries = {{HRI} '25},\n\ttitle = {Online {Prediction} of {User} {Enjoyment} in {Human}-{Robot} {Dialogue} with {LLMs}},\n\turl = {https://dl.acm.org/doi/10.5555/3721488.3721680},\n\tbooktitle = {Proceedings of the 2025 {ACM}/{IEEE} {International} {Conference} on {Human}-{Robot} {Interaction}},\n\tpublisher = {IEEE Press},\n\tauthor = {Janssens, Ruben and Pereira, André and Skantze, Gabriel and Irfan, Bahar and Belpaeme, Tony},\n\tyear = {2025},\n\tpages = {1363--1367},\n}\n\n\n\n\n
\n\n\n\n
\n\n\n
\n
\n\n \n \n \n \n \n \n Between You and Me: Ethics of Self-Disclosure in Human-Robot Interaction.\n \n \n \n \n\n\n \n Irfan, B.; and Skantze, G.\n\n\n \n\n\n\n In
Proceedings of the 2025 ACM/IEEE International Conference on Human-Robot Interaction, of
HRI '25, pages 1357–1362, Melbourne, Australia, 2025. IEEE Press\n
\n\n
\n\n
\n\n
\n\n \n \n
Paper\n \n \n\n \n\n \n link\n \n \n\n bibtex\n \n\n \n\n \n \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@inproceedings{irfan_between_2025,\n\taddress = {Melbourne, Australia},\n\tseries = {{HRI} '25},\n\ttitle = {Between {You} and {Me}: {Ethics} of {Self}-{Disclosure} in {Human}-{Robot} {Interaction}},\n\turl = {https://dl.acm.org/doi/10.5555/3721488.3721679},\n\tbooktitle = {Proceedings of the 2025 {ACM}/{IEEE} {International} {Conference} on {Human}-{Robot} {Interaction}},\n\tpublisher = {IEEE Press},\n\tauthor = {Irfan, Bahar and Skantze, Gabriel},\n\tyear = {2025},\n\tpages = {1357--1362},\n}\n\n\n\n\n
\n\n\n\n
\n\n\n
\n
\n\n \n \n \n \n \n \n Applying General Turn-taking Models to Conversational Human-Robot Interaction.\n \n \n \n \n\n\n \n Skantze, G.; and Irfan, B.\n\n\n \n\n\n\n In
Proceedings of the 2025 ACM/IEEE International Conference on Human-Robot Interaction, of
HRI '25, pages 859–868, Melbourne, Australia, 2025. IEEE Press\n
\n\n
\n\n
\n\n
\n\n \n \n
Paper\n \n \n\n \n\n \n link\n \n \n\n bibtex\n \n\n \n\n \n \n \n 4 downloads\n \n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@inproceedings{skantze_applying_2025,\n\taddress = {Melbourne, Australia},\n\tseries = {{HRI} '25},\n\ttitle = {Applying {General} {Turn}-taking {Models} to {Conversational} {Human}-{Robot} {Interaction}},\n\turl = {https://dl.acm.org/doi/10.5555/3721488.3721593},\n\tbooktitle = {Proceedings of the 2025 {ACM}/{IEEE} {International} {Conference} on {Human}-{Robot} {Interaction}},\n\tpublisher = {IEEE Press},\n\tauthor = {Skantze, Gabriel and Irfan, Bahar},\n\tyear = {2025},\n\tpages = {859--868},\n}\n\n\n\n\n
\n\n\n\n
\n\n\n
\n
\n\n \n \n \n \n \n \n Between Reality and Delusion: Challenges of Applying Large Language Models to Companion Robots for Open-Domain Dialogues with Older Adults.\n \n \n \n \n\n\n \n Irfan, B.; Kuoppamäki, S.; Hosseini, A.; and Skantze, G.\n\n\n \n\n\n\n
Autonomous Robots, 49(9). 2025.\n
Publisher: Springer\n\n
\n\n
\n\n
\n\n \n \n
Paper\n \n \n\n \n \n doi\n \n \n\n \n link\n \n \n\n bibtex\n \n\n \n\n \n \n \n 3 downloads\n \n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@article{irfan_between_2025,\n\ttitle = {Between {Reality} and {Delusion}: {Challenges} of {Applying} {Large} {Language} {Models} to {Companion} {Robots} for {Open}-{Domain} {Dialogues} with {Older} {Adults}},\n\tvolume = {49},\n\turl = {https://rdcu.be/ecPge},\n\tdoi = {10.1007/s10514-025-10190-y},\n\tnumber = {9},\n\tjournal = {Autonomous Robots},\n\tauthor = {Irfan, Bahar and Kuoppamäki, Sanna and Hosseini, Aida and Skantze, Gabriel},\n\tyear = {2025},\n\tnote = {Publisher: Springer},\n}\n\n
\n\n\n\n
\n\n\n\n\n\n