8 entries
Proposes a spectrum of interface metaphors for LLMs beyond the default anthropomorphic framing. Different metaphors (tool, oracle, collaborator, environment) carry different implications for user expectations and design.
Proposes a paradigm where AI functions as a continuously thinking entity rather than a passive responder, enabling proactive collaboration through surfacing intermediate thoughts, full-duplex processing, and shared cognitive workspaces.
A CHI workshop paper proposing tangible interaction metaphors as alternatives to the chat paradigm. Physical and spatial interaction models offer richer interpretative possibilities than conversational turn-taking.
Argues that machines designed around planning models miss the inherently situated nature of human communication. Actions emerge from context and interaction, not from pre-formed plans.
Proposes treating thought itself as the primary substrate for human-AI interaction. Rather than organizing interfaces around conversational turn-taking, design should be built around cognitive processes: attention, intention, reflection.
Synthesises findings from 34 papers on how generative AI expands automation while simultaneously creating risks and opportunities for human cognition. Covers memory, creativity, and critical thinking as the three main domains affected.
VUIs should not commit to a single metaphor but dynamically shift their framing based on context: sometimes assistant, sometimes tool, sometimes ambient presence.
A collection of IBM Research studies on real-world conversational UX, examining how people actually interact with dialogue systems in deployed contexts