När robotar blir karaktärer : Integration av personlighetsdrag och narrativa element i sociala guiderobotar genom värderingsdriven affekt

Read full text Summarise with VIRAT
Authors: Bella Engelson
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This thesis explores how social guide robots can be designed to be perceived as coherent characters by combining trait based personality configuration with narrative framing and appraisal-driven affect modelling. An affect pipeline was implemented on the Roboguide platform by Dyno Robotics, running on an Orion Star GreetingBot Mini. Each conversational turn was appraised by an LLM along eleven cognitive dimensions, mapped to both discrete OCC emotions and a continuous Pleasure Arousal Dominance state, and expressed in real time through facial animation, subtle head movement, and affect conditioned dialogue generation. A mixed method user study (N=20) compared two configurations with identical trait and affect mechanisms but different framing: one presented the robot as a generic guide and one provided a backstory and identity cues. Results suggest that, in this sample, narrative framing was associated with more exploratory and relational interaction and with slightly higher perceived safety. Perceived Big Five traits also tended to shift toward somewhat higher alignment with the configured levels. Alignment between perceived affect and the internal state appeared strongest for pleasure and weaker for arousal and dominance, which may indicate that additional expressive channels are needed for stronger legibility. The work contributes an operationalised affect pipeline and empirical insight into how narrative context shapes robot character perception.

Fields of Study & Subjects

Social Robot Interaction and HRI Speech and dialogue systems Emotion and Mood Recognition

Publication details

Journal / source
KTH Publication Database DiVA (KTH Royal Institute of Technology)
Publication date
1 January 2026
Language
EN
Document type
Thesis