IJCAI is a major international conference on artificial intelligence. Its main goal is to promote sharing (through its annual physical meeting) and dissemination (through its archival proceedings) of knowledge produced by research on artificial intelligence. As a consequence, the ethics of AI is an important concern for IJCAI. We believe that IJCAI publications should not only present high-quality research results, but should also follow the ethical principles of the conference.

Evaluating ethical issues in our field is a challenging task. Most papers focus on general-purpose methodologies.  The real-world impact of a contribution often emerges from the cumulative progress of several papers. It is thus difficult to attribute the impact to an individual paper. Ethical concerns are more apparent when considering applications ready to be deployed.

IJCAI authors are invited to think about the potential negative societal impacts of their contribution.  Whenever a submission is associated with direct potential negative impacts (e.g., the contribution is an application ready for practical use with potential misuses, or the core contribution is inseparable from a questionable application), it should include a discussion of these impacts as well as possible mitigation strategies.  In certain cases, a contribution can have both significant risks and benefits, or it may not be possible to draw a bright line between ethical and unethical. In such cases, authors should not hesitate to critically examine whether the benefits outweigh the potential risks, or to acknowledge ethical ambiguities.

To determine whether a contribution has potential negative societal impacts, consider whether the contribution can directly facilitate injury to living beings, can raise safety, privacy, or security concerns, can raise human rights concerns, can develop or extend harmful forms of surveillance, can severely damage the environment, or can deceive people in ways that cause harm.

If the research uses human-derived data, consider whether that data might:

1. contain personally identifiable information, or information that could be deduced about individuals that they have not consented to share.

2. contain, or potentially exacerbate bias against people of a certain gender, race, sexuality.

3. be obtained via human subject experimentation, and, if so, whether the research has been reviewed and approved by a relevant oversight board.

Other issues related to data include consent to use or share the data. If you did not receive consent, explain why this might be appropriate from an ethical standpoint.

These lists are not intended to be exhaustive —they are included here as a prompt for author and reviewer reflection.

During the IJCAI review process, reviewers are asked to flag papers with significant ethical concerns. These will be referred to an ethics committee, which will assess the situation and advise the program chair. The program chair reserves the right to reject papers with grave ethical issues, but expects this to occur only in exceptional circumstances.