Plan, Activity and Intent Recognition (PAIR)

AAAI-26 Tutorial • Tuesday, January 20 • 8:30am–12:30pm

Location: Singapore EXPO Convention & Exhibition Centre

AAAI-26 Half-day Tutorial 8:30am – 12:30pm January 20, 2026 Singapore EXPO

About the Tutorial

Despite rapid progress in machine learning, real-time, online inference of agents’ intentions remains one of the enduring grand challenges in AI. Plan, activity, intent, and goal recognition focus on inferring what software agents, robots, or humans are trying to achieve from observations of their behavior and interactions.

This tutorial targets AI students and researchers who want to understand and apply recognition techniques in their own work, and who are looking for new research directions grounded in this rich, cross-cutting area of AI.

Goal
Provide a unified, up-to-date introduction to plan, activity, intent, and goal recognition: core concepts, problem formulations, algorithms, and applications.
Who should attend?
Graduate students, researchers, and practitioners with interest in planning, reinforcement learning, human–robot interaction, multi-agent systems, or intent inference.
Prerequisites
Familiarity with basic AI concepts, probabilistic reasoning, and planning is recommended. Prior exposure to Bayesian inference or logic-based reasoning is helpful but not required.

Motivation & Scope

Recognition problems arise in a wide range of domains: assistive technologies, software assistants, computer and network security, behavior recognition, human–robot collaboration, and more. Techniques span user modeling, machine vision, automated planning, intelligent user interfaces, HCI, multi-agent systems, natural language understanding, and machine learning.

While this diversity has produced a wealth of ideas and tools, it has also led to fragmentation. This tutorial brings together perspectives and methods from multiple AI subfields to offer a coherent view of the plan, activity, intent, and goal recognition landscape.

Building on the successful AAAI-19 PAIR tutorial and our 2021 book, we present an updated and expanded view of the field.

Keywords

Plan recognition Goal recognition Activity recognition Behavior recognition Temporal pattern recognition Probabilistic inference Reasoning under uncertainty Theory of mind Human–AI collaboration Multi-agent systems

Detailed Outline

  1. Introduction to recognition problems
    Plan, activity, intent, and goal recognition: basic concepts and problem space.
  2. Motivating scenarios across domains
    Examples from human–robot collaboration, security, assistive technologies, and other application areas.
  3. Formal problem definitions
    Distinctions among plan, activity, intent, and goal recognition tasks; relationships between them.
  4. Computational representations
    Logic-based, probabilistic, machine learning, and hybrid models used for recognition.
  5. Algorithms and inference methods
    Core approaches for performing online and offline recognition in different settings.
  6. Evaluation methodologies and benchmarks
    How to evaluate recognition systems: datasets, metrics, and experimental design.
  7. Current challenges & open directions
    Scalability, real-time reasoning, robustness, integration with learning, human–AI collaboration, and other open research questions.

Schedule Breakdown

The tutorial is scheduled for Tuesday, January 20, 8:30am–12:30pm at the Singapore EXPO Convention & Exhibition Centre.

  • 08:30 – 08:40  Welcome, logistics, and motivation
  • 08:40 – 09:20  Introduction to plan, activity, intent, and goal recognition
  • 09:20 – 10:00  Formal problem definitions and core representations
  • 10:00 – 10:30  Break
  • 10:30 – 11:10  Algorithms and inference methods (online & offline)
  • 11:10 – 11:50  Applications, evaluation, and benchmarks
  • 11:50 – 12:30  Open challenges, discussion, and Q&A

Times are approximate and may be adjusted slightly during the tutorial to allow for questions and discussion.

Presenters

Sarah Keren
Sarah Keren
Senior Lecturer, Faculty of Computer Science, Technion – Israel Institute of Technology
Contact: sarahk@technion.ac.il

Sarah’s research focuses on multi-agent environment design: creating environments that enhance agent capabilities and enable effective multi-robot and human–robot collaboration. She combines model-based reasoning, decision-making under uncertainty, game theory, and multi-agent learning. Her doctoral work introduced Goal Recognition Design, a framework for improving goal inference via environment redesign.

Reuth Mirsky
Reuth Mirsky
Assistant Professor, Computer Science Department, Tufts University
Contact: reuth.mirsky@tufts.edu

Reuth leads the GOLD (goal optimization using learning and decision-making) lab. Her work develops algorithms and frameworks that challenge traditional assumptions about intelligent agents. She is an active contributor to the AI and HRI communities and has co-organized more than 15 events on related topics at major AI conferences, including the PAIR and RaD-AI workshop series.

Christopher Geib
Christopher Geib
Principal Scientist, Charles River Analytics
Contact: cwgeib@gmail.com

Chris is the principal architect of multiple probabilistic plan recognition systems, including the ELEXIR system, which demonstrates state-of-the-art plan recognition and planning capabilities using a single, shared, and learnable domain representation.

Materials

We will make tutorial materials available here for participants before and after the event.

References

  • S. Keren, R. Mirsky, and C. Geib. “Plan Activity and Intent Recognition Tutorial.” AAAI, 2019.
  • R. Mirsky, S. Keren, and C. Geib. Introduction to Symbolic Plan and Goal Recognition. Springer, 2021.