RW
R. Weijers
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
5 records found
1
When Should Robots Speak?
The Effect of Proactive vs. Reactive Robot Interventions on Perceived Autonomy in Creative Group Ideation
This study investigates whether proactive robot interventions undermine autonomy compared to reactive interventions during collaborative ideation tasks. A Pepper robot was implemented as a conversational agent capable of generating spoken solution suggestions using a locally hosted large language model, speech recognition, and text-to-speech functionality. The study used a between-subjects design with two conditions: a proactive condition, in which the robot autonomously intervenes based on conversational cues such as silence or signs of struggle, and a reactive condition, in which it only responds when explicitly addressed. Participants completed a collaborative task focused on improving campus life, after which perceived autonomy was measured using an adapted autonomy subscale of the Basic Psychological Need Satisfaction and Frustration Scale. Behavioral data, including robot intervention frequency, timing, and number of ideas generated, were also collected. Results showed no statistically significant differences between conditions, suggesting that proactive interventions do not substantially reduce participants’ sense of autonomy. Groups in the proactive condition generated an average of 7.70 ideas compared to 5.80 in the reactive condition, but this difference was not significant (p = .067). These findings suggest that unsolicited robot participation may not inherently undermine perceived autonomy in short brainstorming tasks, while highlighting the importance of intervention design.
...
This study investigates whether proactive robot interventions undermine autonomy compared to reactive interventions during collaborative ideation tasks. A Pepper robot was implemented as a conversational agent capable of generating spoken solution suggestions using a locally hosted large language model, speech recognition, and text-to-speech functionality. The study used a between-subjects design with two conditions: a proactive condition, in which the robot autonomously intervenes based on conversational cues such as silence or signs of struggle, and a reactive condition, in which it only responds when explicitly addressed. Participants completed a collaborative task focused on improving campus life, after which perceived autonomy was measured using an adapted autonomy subscale of the Basic Psychological Need Satisfaction and Frustration Scale. Behavioral data, including robot intervention frequency, timing, and number of ideas generated, were also collected. Results showed no statistically significant differences between conditions, suggesting that proactive interventions do not substantially reduce participants’ sense of autonomy. Groups in the proactive condition generated an average of 7.70 ideas compared to 5.80 in the reactive condition, but this difference was not significant (p = .067). These findings suggest that unsolicited robot participation may not inherently undermine perceived autonomy in short brainstorming tasks, while highlighting the importance of intervention design.
Asking the Right Question
How Robot Elicitation Strategies Shape Engagement and Substantive Contribution in Creative Group Ideation
Social robots can shape group interaction, but robot facilitation is often studied at the level of the robot as a whole rather than at the level of the specific utterances through which the robot intervenes. This paper investigates how three spoken elicitation strategies delivered by a social robot -generative, elaborative, and perspective-shifting prompts- shape participant engagement and contribution substantiveness in a two-person creative ideation task about improving the TU Delft campus experience. The study used a within-subjects design in which each of the 20 groups received one prompt from each strategy during divergence and one during convergence, while the task, robot, scheduling rule, and technical setup were kept constant. Engagement was measured using self-reports, response delay, speaking time, vocal activation, and connection cues; contribution substantiveness was measured using manually coded idea count, elaboration units, and consecutive same-subject turns. The descriptive results suggest that convergence produced higher self-reported engagement than divergence, generative prompts were most associated with idea breadth, and elaborative prompts were most associated with developed and sustained discussion. Perspective-shifting prompts appeared more useful once participants already had ideas to evaluate. The findings do not establish a universally superior strategy, but they show that robot facilitation should be designed as phase-sensitive prompt behaviour rather than treated as a general effect of robot presence.
...
Social robots can shape group interaction, but robot facilitation is often studied at the level of the robot as a whole rather than at the level of the specific utterances through which the robot intervenes. This paper investigates how three spoken elicitation strategies delivered by a social robot -generative, elaborative, and perspective-shifting prompts- shape participant engagement and contribution substantiveness in a two-person creative ideation task about improving the TU Delft campus experience. The study used a within-subjects design in which each of the 20 groups received one prompt from each strategy during divergence and one during convergence, while the task, robot, scheduling rule, and technical setup were kept constant. Engagement was measured using self-reports, response delay, speaking time, vocal activation, and connection cues; contribution substantiveness was measured using manually coded idea count, elaboration units, and consecutive same-subject turns. The descriptive results suggest that convergence produced higher self-reported engagement than divergence, generative prompts were most associated with idea breadth, and elaborative prompts were most associated with developed and sustained discussion. Perspective-shifting prompts appeared more useful once participants already had ideas to evaluate. The findings do not establish a universally superior strategy, but they show that robot facilitation should be designed as phase-sensitive prompt behaviour rather than treated as a general effect of robot presence.
When Robots Brainstorm With Us
Robot Facilitation and Social Comparison in Creative Group Ideation
Social robots are increasingly designed to contribute to human tasks, but their competence may also affect how users evaluate their own abilities. This study investigated how different Pepper facilitation configurations shape social comparison, self-efficacy, perceived competence, and perceived contribution during creative group ideation. A between-subjects study was conducted with 20 participants in 10 dyads. Participants brainstormed ideas for improving campus wellbeing and academic engagement with Pepper in one of two conditions: a Dynamic Collaborative Pepper condition, in which Pepper responded to the discussion in real time, or a Pre-Generated Intervention Pepper condition, in which Pepper contributed prepared ideas and visual material at planned moments. Descriptive results suggest that the dynamic collaborative condition was experienced as more collaborative and was associated with higher self-efficacy and perceived competence. The pre-generated intervention condition was perceived as slightly more useful and engaging, but produced stronger social comparison and made Pepper appear more influential in the final solution. These findings suggest a design trade-off: polished robot contributions may support the task while increasing upward comparison, whereas adaptive responses generated during the discussion may better preserve participants' sense of agency and competence.
...
Social robots are increasingly designed to contribute to human tasks, but their competence may also affect how users evaluate their own abilities. This study investigated how different Pepper facilitation configurations shape social comparison, self-efficacy, perceived competence, and perceived contribution during creative group ideation. A between-subjects study was conducted with 20 participants in 10 dyads. Participants brainstormed ideas for improving campus wellbeing and academic engagement with Pepper in one of two conditions: a Dynamic Collaborative Pepper condition, in which Pepper responded to the discussion in real time, or a Pre-Generated Intervention Pepper condition, in which Pepper contributed prepared ideas and visual material at planned moments. Descriptive results suggest that the dynamic collaborative condition was experienced as more collaborative and was associated with higher self-efficacy and perceived competence. The pre-generated intervention condition was perceived as slightly more useful and engaging, but produced stronger social comparison and made Pepper appear more influential in the final solution. These findings suggest a design trade-off: polished robot contributions may support the task while increasing upward comparison, whereas adaptive responses generated during the discussion may better preserve participants' sense of agency and competence.
Social robots deployed in collaborative groups reshape interaction between human participants, yet how different verbal strategies produce distinct shaping effects remains untested. This study compared assertive and supportive robot strategies across 20 dyads in a creative ideation task with a Pepper robot. Transcripts were coded using Mercer’s talk taxonomy and post-session interviews were analysed thematically. The supportive strategy produced a significantly higher frequency of exploratory discourse than the assertive strategy. The interaction strategy had no significant effect on participation balance, disputational talk, cumulative talk, group cohesion, or ingroup identification. Four unintended cross-condition shaping effects emerged, including attention redistribution, partner solidarity, speech formalisation,
and emotional suppression. Condition-specific effects were also observed, including creative suppression and expert deference. These results show that a robot’s verbal register shapes the quality of human collaboration, and that robot presence alone restructures interaction beyond any programmed design intent. ...
and emotional suppression. Condition-specific effects were also observed, including creative suppression and expert deference. These results show that a robot’s verbal register shapes the quality of human collaboration, and that robot presence alone restructures interaction beyond any programmed design intent. ...
Social robots deployed in collaborative groups reshape interaction between human participants, yet how different verbal strategies produce distinct shaping effects remains untested. This study compared assertive and supportive robot strategies across 20 dyads in a creative ideation task with a Pepper robot. Transcripts were coded using Mercer’s talk taxonomy and post-session interviews were analysed thematically. The supportive strategy produced a significantly higher frequency of exploratory discourse than the assertive strategy. The interaction strategy had no significant effect on participation balance, disputational talk, cumulative talk, group cohesion, or ingroup identification. Four unintended cross-condition shaping effects emerged, including attention redistribution, partner solidarity, speech formalisation,
and emotional suppression. Condition-specific effects were also observed, including creative suppression and expert deference. These results show that a robot’s verbal register shapes the quality of human collaboration, and that robot presence alone restructures interaction beyond any programmed design intent.
and emotional suppression. Condition-specific effects were also observed, including creative suppression and expert deference. These results show that a robot’s verbal register shapes the quality of human collaboration, and that robot presence alone restructures interaction beyond any programmed design intent.
As artificial intelligence transitions into physically embodied collaborative roles, systems must be designed to support rather than thwart human autonomy. In a between-subjects experiment (N = 20 dyads), we investigated how the conversational delivery style (Assertive vs. Supportive) of an LLM-driven Pepper robot affected perceived user autonomy and creative output during a live campus-improvement brainstorming task. While both conditions yielded a similar quantity of unique ideas (Assertive M = 5.90, Supportive M = 5.70, t(18) = 0.40, p = 0.697), the Supportive delivery style marginally increased autonomy frustration (M = 2.43) compared to the Assertive style (M = 2.03; t(38) = 2.02, p = 0.050, d = 0.64). Qualitative analysis of the interaction logs revealed that the Supportive robot's conversational pacing frequently triggered interruptions during natural cognitive silences, actively seizing the conversational floor and disrupting the dyads' workflow. These findings demonstrate that in live human-robot interaction, the structural implementation of interaction timing and turn-taking is more critical to preserving perceived user autonomy than empathetic semantic framing.
...
As artificial intelligence transitions into physically embodied collaborative roles, systems must be designed to support rather than thwart human autonomy. In a between-subjects experiment (N = 20 dyads), we investigated how the conversational delivery style (Assertive vs. Supportive) of an LLM-driven Pepper robot affected perceived user autonomy and creative output during a live campus-improvement brainstorming task. While both conditions yielded a similar quantity of unique ideas (Assertive M = 5.90, Supportive M = 5.70, t(18) = 0.40, p = 0.697), the Supportive delivery style marginally increased autonomy frustration (M = 2.43) compared to the Assertive style (M = 2.03; t(38) = 2.02, p = 0.050, d = 0.64). Qualitative analysis of the interaction logs revealed that the Supportive robot's conversational pacing frequently triggered interruptions during natural cognitive silences, actively seizing the conversational floor and disrupting the dyads' workflow. These findings demonstrate that in live human-robot interaction, the structural implementation of interaction timing and turn-taking is more critical to preserving perceived user autonomy than empathetic semantic framing.