Elements of Task Theory

Matteo Belenchia*, Kristinn R. Thórisson, Leonard M. Eberding, Arash Sheikhlar

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review


Tasks are of primary importance for artificial intelligence (AI), yet no theory about their characteristics exists. The kind of task theory we envision is one that allows an objective comparison of tasks, based on measurable physical properties, and that can serve as a foundation for studying, evaluating, and comparing learning controllers of various kinds on a variety of tasks by providing principled ways for constructing, comparing, and changing tasks with particular properties and levels of difficulty. In prior papers we have outlined an approach towards this goal; in this paper we present further principles for its development, including causal relations. We use these principles to expand our prior ideas, with the aim of laying the groundwork for covering levels of detail, prior knowledge of the learner/performer and task difficulty, to name some of the complex issues that must be solved for a useful task theory.

Original languageEnglish
Title of host publicationArtificial General Intelligence - 14th International Conference, AGI 2021, Proceedings
EditorsBen Goertzel, Matthew Iklé, Alexey Potapov
PublisherSpringer Science and Business Media Deutschland GmbH
Number of pages11
ISBN (Print)9783030937577
Publication statusPublished - 2022
Event14th International Conference on Artificial General Intelligence, AGI 2021 - San Francisco, United States
Duration: 15 Oct 202118 Oct 2021

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13154 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Conference14th International Conference on Artificial General Intelligence, AGI 2021
Country/TerritoryUnited States
CitySan Francisco

Bibliographical note

Publisher Copyright:
© 2022, Springer Nature Switzerland AG.

Other keywords

  • Artificial intelligence
  • Environments
  • Evaluation
  • General machine intelligence
  • Task theory
  • Tasks


Dive into the research topics of 'Elements of Task Theory'. Together they form a unique fingerprint.

Cite this