W3C

– DRAFT –
Multi-Agent Conversational Methodology Breakout

25 February 2026

Attendees

Present
Deborah_Dahl, Emmett_Coin, Kaz_Ashimura, Matt_Shomphe, Molly_Lewis, plh, Samantha_Estoesta
Regrets
-
Chair
-
Scribe
kaz

Meeting minutes

Self Intro

(everybody gives their intro)

Discussion

se: different cultural groups use agents
… different communications styles with agents

ec: great question
… e.g., Chinese agents vs Australian agents

ms: trying to negotiate cultural differences
… if two agents talking with each other, e.g., one from China and another from Australia

ec: might be intentional
… natural to happen
… also should think about agents working with people
… maybe we should have some protocol to handle the situation

se: other issue about global languages
… could have shared common language dictionary
… to handle difficulty cultures and styles

ec: interesting
… e.g., English as the second language
… we could have a standard like English language
… kind of like Esperanto

ms: think so too

ec: some way to translate to a standard
… kind of like multilingual
… for example, would be useful to have an agent for warehouse coordination

ms: can we imagine shipping products, e.g., to Ontario
… you have multi-lingual environment
… like the idea somebody talking based on the context
… there are many ways to think about that
… some sort of reduced English
… translated from other languages
… e.g., from Turkish

ec: like Star Trek :)

ms: right
… universal translator
… multi-agent and multi-languages

ec: maybe agents should handle language modules on top of the structure
… very structured language agents could use
… mostly "language" is important for humans

kaz: should think about both NL and structured model
… also we should think about dialog management model to handle multiple agents and humans

ec: right
… have to have similar thing
… typical agents inside our standard field
… we could add the model why the human is raising hand

kaz: right

(Debbie jois)

ec: (gives summary so far)
… should think about difference based on language, culture, etc.
… for example, there are girls who need help, and agents can help them go home
… it's kind of like a game

ms: interesting
… customers would like to achieve some goal

ec: when we generate an LLM, should get landmarks, etc.

se: do we have to build to understand agents?

ec: very helpful rules can be built

dd: plh reminded me that there is just 10 more minutes

ec: ok. any more things?

dd: discussion around cultural issues?

ec: yeah
… had discussion about cultural differences
… mid-west vs south America, etc.

ms: impressive

ec: agents like actors from Shakespeare
… how we signal cultural differences?

se: so many disclosures
… languages we use for training
… English, e.g., American English
… need a good standard for that

dd: some format for that purpose

ec: have strong attitude for that
… what the agent is like
… not necessarily behaving right to people

ms: right
… wouldn't trust developers accurately express it

ec: possibly, we could have agents which have different purposes/attitudes

ms: yeah, could be malicious

[breakout ends]

Minutes manually created (not a transcript), formatted by scribe.perl version 248 (Mon Oct 27 20:04:16 2025 UTC).

Diagnostics

Succeeded: s/rant/ranto/

Succeeded: s/help/help, and agents can help them go home/

No scribenick or scribe found. Guessed: kaz

Maybe present: dd, ec, kaz, ms, se

All speakers: dd, ec, kaz, ms, se

Active on IRC: kaz