In recent decades, human-robot interaction (HRI) has evolved from physical manipulation to information exchange and socialisation. Although the use of speech-enabled social robots is growing in many ways, the improvement of the HRI experience remains limited. Users are often disappointed after the novelty of the first encounter fades.
Admittedly, users' expectations of social robots are influenced by science fiction and selectively presented advertising. However, another important factor deserves attention: the design of the robot's look, sound, and interaction behaviour, known as its 'affordance'. Appropriate affordance design helps users form realistic expectations about what an artefact can do and how it should be used. On the one hand, humanlike affordances are popular and may initially foster engagement. On the other hand, such designs bear risks: they can evoke discomfort, trigger the 'uncanny valley' effect, or raise ethical concerns about misleading users.
Beyond enhancing robots' interaction abilities, it is therefore crucial to investigate how affordance design can support socially meaningful, trustworthy interaction, especially given that social robots remain fundamentally mismatched conversational partners compared to humans. The work presented here explores the overarching research aim: How can the affordance design of speech-enabled social robots be constructed to support positive and contextually appropriate user experiences? The research adopts an interdisciplinary approach, combining insights from psychology, interaction design, and computer science. Through a series of empirical studies and theoretical developments, it investigates how external interaction cues can be aligned with internal robot capabilities to produce honest, coherent, and context-sensitive affordances.
The thesis title, 'My Fair Robot', draws inspiration from the play 'My Fair Lady', where transformation is not about superficial change alone but about achieving deep, consistent alignment between outward presentation and internal state. Similarly, this research advocates for multimodal, internally consistent, and contextually honest affordance design that helps robots not merely to appear capable, but to transparently and reliably reflect what they can genuinely deliver in human-robot interaction.
