EMCAI Meeting – Friday 27 September – 12.30-14.00 CET

The last EMCAI session outlined several interesting “assumptions” on what constitutes “interaction” in “human-robot interaction” or “human-VUI interaction”. Because the discussion was still lively when time ran out, we invite you to join us for one final session on the very same topic: What do “we” (EMCA researchers, non-EMCA-oriented HRI researchers, engineers, designers, tech companies’ employees, etc.) respectively index when we formulate phrases such as “conversing with a robot”, “interacting with a vocal agent”, etc.?

Outlined below is a tentative list of the prominent assumptions debated last time. Some of those assumptions may prevail in “the industry” (i.e., in private companies working on commercial robots or VUIs), while others may be at the core of different disciplines in HRI/HCI academic research. Some might be documented across multiple papers and fields, while others might still remain unexplored.

1. Human-robot interaction as “information transfer” rather than “dealing with practical problems in situated activities”.
2. Human-robot interaction as questions and answers – a focus on turns’ “composition” over turns’ “position”.
3. Human-robot interaction as non-contingent, definite, and exhaustively describable – the definiteness of reality and the possibility of its description rather than the “essential vagueness” of social life.
4. Human-robot Conversation as a practical achievement that it is pointless to reconstruct analytically (e.g., using CA) to, then, extract a set of granular guidelines or rules – conversational design as “gut feeling”.
5. Human-robot interaction as interactions with machines and not interactions that involve machines – “human-machine coupling” rather than “interaction” in a situated and holistic sense.

1) For the next session: To expand, correct, or refine this brief preliminary list, we invite you to prepare any data that you would like to discuss (articles, videos, documents, etc.) and that may exemplify a common assumption about “human-robot interaction” or “human-VUI interaction”. Our intention is not to have a debate on “interaction” as a concept, but to investigate what is indexed by different HRI actors as an “interaction”, and to explore if these assumptions impact the design of robots or VUIs – as well as the tools used to design or program them. For example, do these assumptions find their way into the most recent conversational technologies, such as chatbots based on large language models like chatGPT? And if so, what evidence can substantiate these claims? We would love to discuss any data you might want to present, even tangentially related to these topics.

2) As a starting point, the candidate assumptions discussed during our previous session have been tentatively mapped onto a Miro board. This Miro board lays out assumptions that were hypothesized to be prevailing in HRI, as well as alternative positions that contrasted with those prevailing assumptions:

[Please e-mail Lynn (lynn.derijk @ ru.nl) for access to the Miro board if you are not yet on our mailing list and thus missed the invitation e-mail]

We will try to use this board as a basis for our discussions next time. In the meanwhile, we invite you to modify this collaborative Miro board as you see fit. Feel free to add, modify, or move assumptions, contrasting assumptions, comments, criticism, questions, empirical data, or to draw new clusters around the preexisting sticky notes. Similarly, do not hesitate to heavily update the Miro board if you think about better organizing principles or if you have more catchy names for existing assumptions. (For those who participated in our last session, sorry in advance if we miscategorized your verbatims!