Maoinithe ag an Aontas Eorpach. Mar sin féin is tuairimí agus tuairimí an údair amháin nó na n-údar amháin iad na tuairimí agus na tuairimí a nochtar agus ní gá go léiríonn siad tuairimí an Aontais Eorpaigh nó na Gníomhaireachta Eorpaí Feidhmiúcháin Oideachais agus Cultúir (EACEA). Ní fhéadfar an tAontas Eorpach ná EACEA a chur i gcuntas astu. Uimhir an tionscadail: 2023-1-NL01-KA220-HED-000155675. Cás-Staidéar 19: Tacaíocht a thabhairt d'Oideachas Eolaíochta Ar Líne Comhoibríoch le Gníomhaire Comhrá Inaistrithe agus Inchumraithe General information Reference/Source: Araujo, A. De, Papadopoulos, P. M., McKenney, S., & Jong, T. De. (2023). Supporting Collaborative Online Science Education with a Transferable and Configurable Conversational Agent. Computer-Supported Collaborative Learning Conference, CSCL, 2023-June, 416–419. https://doi.org/10.22318/cscl2023.469853 Institution: University of Twente (Netherlands) Course/Subject: Science, photosynthesis Aim: To develop and pilot a transferable and configurable conversational agent (Clair) designed to facilitate productive talk in collaborative online learning environments. Target group: Students in pairs within collaborative online learning settings (AI developed to support students learning process) Description of case Overview: Researchers designed a conversational agent named Clair to foster productive talk in collaborative online learning environments. Clair is intended to be transferable to different topics and languages and allow for a degree of teacher configuration. Intervention: The pilot study used a within-subjects experiment. Students worked in pairs on a Go-Lab activity about photosynthesis. After an initial phase without Clair, dyads were assigned to 'control' or 'treatment' groups, with the treatment group receiving Clair's interventions. Clair's design: Clair used talk moves (e.g., Add-on, Rephrasing, Expand Reasoning) to stimulate discussion based on a combination of dialogue variables (focus, intent, topic similarity, etc.) and fuzzy logic rules. Lessons learned Limited impact: While Clair showed some potential in increasing explicit reasoning and decreasing participation imbalance, the overall effect wasn't statistically significant. Design issues: Clair's interventions were perceived as repetitive and robotic. The triggering mechanisms and rules could be improved. Unrealistic expectations: Students expected Clair to provide more direct content support, which isn't its intended function. Key takeaways: Designing an effective conversational agent for collaborative learning is complex. Future iterations should focus on more nuanced interventions, better rule design, and managing student expectations about the CA's capabilities. Implications for practice Practitioners must explicitly define the AI's role as a social facilitator rather than a content expert, while diversifying its "talk moves" to prevent student disengagement from repetitive or robotic interactions.
RkJQdWJsaXNoZXIy NzYwNDE=