Re: Invited Talk: Amit Chopra — Why Agentic AI Needs Interaction-Oriented Programming (Aug 1, 10:15 CEST)

Dear all,

The recording of Amit's talk is now available on YouTube: https://youtu.be/LpAC-sipBvA

The PR for the meeting notes is here: https://github.com/w3c-cg/webagents/pull/90

Best wishes,
Andrei



> On 21 Jul 2025, at 10:46, Andrei Ciortea <andrei.ciortea@inria.fr> wrote:
> 
> Dear CG participants,
> 
> At our next regular meeting on August 1, we're pleased to welcome Prof. Amit K. Chopra (Lancaster University) for an invited talk on multiagent interaction protocols and interaction-oriented programming.
> 
> Please find below the talk abstract. For the meeting details:
> - calendar entry and joining instructions: https://www.w3.org/events/meetings/5a77b997-467b-4d5d-bb40-b84789c17d2e/20250801T090000/
> - date and time according to your location: https://zonestamp.toolforge.org/1754036100
> 
> Please note: To accommodate Amit's schedule, the meeting will begin at *10:15 CEST*, which is 1 hour and 15 minutes later than usual.
> 
> We look forward to seeing you online for an exciting session.
> 
> Best wishes,
> Andrei
> 
> *Title:* Why Agentic AI Needs Interaction-Oriented Programming
> 
> *Abstract:* Agentic AI envisages the use of Generative AI, especially LLMs, as a decision-making engine for agents. Its value arises from the ability to reason flexibly in a variety of contexts but without explicit knowledge engineering.  Exploiting this flexibility requires approaches for modeling multiagent interactions that support flexibility. Workflow, the currently dominant paradigm for modeling interactions in Agentic, doesn't fit the bill; it hasn't since the 1990s.
> 
> The multiagent systems community took an alternative approach: With of aim of supporting intelligent decision making by agents, they sought to capture the meaning of interactions. KQML and the FIPA ACL were early missteps. Interaction-Oriented Programming (IOP) addresses these missteps.  IOP is formal, declarative, and general purpose (can be used to model diverse applications). It is motivated from meaning and supports maximally flexible interactions between agents. No approach supports loose-coupling, including asynchrony and heterogeneity, better than IOP. Moreover, IOP boasts a growing repository of software, including compilers, verifiers, and programming models (https://gitlab.com/masr).
> 
> In this talk, I will touch upon the past, the present, and the future of multiagent interactions, as relevant to Agentic AI. Let us not repeat the mistakes of the past.
> 
> *Biography:* Amit K. Chopra is a senior lecturer at Lancaster University in the UK. He is interested in software abstractions that enable engineering flexible, decentralized multiagent systems. The notions of protocols and norms are central to this work, which has been published in prestigious AI conferences such as AAMAS, IJCAI, and AAAI.  
> 
> Amit has given several tutorials and invited talks. His research has been funded by the UKRI and the EU. He has a PhD in Computer Science from North Carolina State University.
> 
> 
> --
> Andrei Ciortea
> 
> Assistant Professor
> School of Computer Science
> University of St.Gallen, Switzerland
> https://interactions.ics.unisg.ch/
> 
> External Collaborator
> Wimmics
> Inria, Université Côte d’Azur, CNRS, I3S, France
> https://wimmics.inria.fr/
>