The Third Annual Conference on Advances in Cognitive Systems
May 28 - 31, Atlanta, GA
Goal Reasoning (GR) is the study of cognitive systems that can dynamically deliberate on, generate, and select their goals. This can be valuable in dynamic environments in which not all events can be predicted or their responses pre-encoded. Under other names, GR has been studied in AI (e.g., intelligent agents, cognitive architectures, planning, machine learning), robotics, cognitive science, and philosophy, among other disciplines. A GR system may include, for example: models of its intentions (e.g., its own drives, preferences, or decision making); models of other agents and their goals/intent; models of exogenous events; a model of attention; a method for interpreting its observations; a method for recognizing objects and events among these observations that require deliberation; a method for interpreting these observations; a decision-making method for selecting post-deliberative responses; a method for searching through, ranking, selecting from, and managing a space of goals in a variety of refinement modes; planners and schedulers to refine goals; and an ability to execute actions associated with selected plans of selected goals. GR has been investigated in domains such as game AI, unmanned autonomous vehicles, cognitive robotics, and intelligent decision aids. Recent GR research has focused on topics such as plan recognition, reasoning about state expectations, event diagnosis, goal hierarchies, goal selection, and goal refinement. However, this has only scratched the surface. This workshop’s objective is to provide an opportunity to present, learn about, and discuss recent GR research and future directions.
Topics relevant to this workshop include, but are not limited to, the following areas:
- Theoretical models of goal reasoning (GR)
- Motivated agents
- Interactive GR
- Goal reasoning in humans
- Constraining GR (e.g., using narrative structures)
- Explanation and diagnosis of notable objects and events
- Planning, scheduling, and (meta-)reasoning with goals
- Multiagent GR systems
- Machine learning for GR
- Comparisons of GR with other models of autonomy
- Evaluation/analyses of goal reasoning
- Demonstrations of GR systems
- Applications (e.g., Game AI, robotics, decision aids, or other domains)
- 13 March - 24 April 2015 (optional): Authors can submit drafts (1-2 pages) for fast-track decision process
- 24 April 2015: Full submissions due
- 4 May 2015: Responses to author submissions
- 11 May 2015: Final versions due
- 14 May 2015: Papers and agenda placed online
- 28 May 2015: Workshop
Although this one-day workshop will include (oral and poster) presentations of selected submissions (and their common themes), we will emphasize invited presentations (and discussion) on topics such as (but not limited to) (1) an overview of existing approaches, (2) theoretical models of GR, (3) GR process components (e.g., models of attention, explanation, leveraging of contextual constraints, goal precedence revision), and (4) the implications of specific application environments (e.g., real-time, multiagent, adversarial), and contexts (e.g., robotics, cognitive architectures, game AI). Discussion time periods will be reserved to provide additional feedback on workshop papers. Interested and curious researchers are most welcome!
This one-day workshop will be held on 28 May 2015 as part of the Advances in Cognitive Systems (ACS-15) conference (28-31 May) in Atlanta, Georgia. This workshop is open to researchers and practitioners interested in GR and related topics. Attendees must register for the ACS conference. (A workshop-only registration option does not exist; see the ACS-15 web site above for details.) If you are a student considering participating in this workshop, Please contact the workshop chair if you have any questions.
To provide authors with sufficient time to make travel plans, we offer an optional fast-track decision process on submissions; authors are welcome to email 1-2 page early versions of their submission to the workshop chair, who will quickly check for workshop relevance and reply on whether they will be accepted. Thus, for completed submissions that were pre-accepted, the traditional reviewing process will focus only on providing useful feedback rather than accept/reject decisions. All other submissions will be reviewed using the normal OC process. Completed paper submissions should be formatted according to the ACS Conference Instructions and submitted using EasyChair; please include your name(s), affiliation(s), and email address(es) at the top of the first page. Submissions should be max 16 pages and in PDF format. We also welcome shorter (max 8 pages) submissions, including on system demonstrations.