Call for Papers, Challenges, and retrospectives

We invite researchers to submit their recent work on the topic in the intersection of Artificial Intelligence (AI) and Humanitarian Assistance and Disaster Recovery (HADR) problems to be featured at our workshop co-located with NeurIPS 2021. This year, we are opening three calls: papers, retrospectives, and challenges.

1. Call for Papers

We welcome methods from Machine Learning, Computer Vision, Robotics, Natural Language Processing, Information Theory, Decision Theory, Game Theory, Causal Modeling, Applied Statistics, and related fields.

We highly encourage work with clear applications to any of a variety of humanitarian and disaster response issues, including, but not limited to, search and rescue robotics, computer vision for damage assessment, motion planning in obstructed roads, response to the COVID-19 pandemic, disinformation management on social media, and much more.

Paper Submission Format

Paper submissions will be limited to five pages of content and one page of references in the NeurIPS 2021 format.  All paper submissions must include an "Application Context" section. This section should discuss why this problem is a pressing concern within the HADR domain, especially specifics as to disaster, crisis, or type of assistance, why your research is impactful in this context, partnerships with agencies or stakeholders, and explain any concepts that would not be common knowledge to a general machine learning researcher (citations from HADR literature encouraged).

Please use your best judgement for what is suitable for your submission. The intent of this section is to increase readability of papers for all workshop participants, from ML experts to domain practitioners of HADR, as well as to help us match your submission to reviewers. 

Papers may contain appendices after the sixth page to highlight additional results, figures, proofs, and other supplementary materials, but the main body of the work should be self-contained.  Other supplementary materials, such as videos or audio clips, may be submitted as supplementary materials.  Reviewers are not required to consider appendices or supplementary materials in their decisions.

Paper Submission Instructions

The deadline for submission is September 25th, 2021 at 11:59PM AoE (anywhere on Earth). Submissions must be made on the hosted CMT instance at https://cmt3.research.microsoft.com/AIHADRW2021/.

All submissions will undergo double-blind review; Thus, please ensure that your submission is anonymized when submitted, including removing names, affiliations, contact information, and ensuring that references to your own previous work are in the third person.

Paper Review Criteria

Paper submissions will be reviewed along the following criteria:

• Clarity in the explanation of the problem setting, including (a) how it relates to the HADR application themes, and (b) why an AI-based approach is suitable and ethical in this domain. For all papers this should be addressed at minimum in the "Application Context" section, see above.

• Technical merit, including methodology, experimental results, and evaluation procedure

• Relevance of the AI components of the work to the intended HADR problem setting(s)

• Demonstrated evidence of or potential for cross-disciplinary takeaways (where applicable)

Each submission will be reviewed by at least one member each from the AI and HADR communities to ensure both technical merit and impact are both considered.

2. Call for Retrospectives

This year we are excited to open a call for "Retrospectives":  submissions that detail AI + HADR research that has seen prototypical or production use in the field, including their points of success and modes of failure. These perspectives are invaluable in learning from past mistakes as well as successes, and for guiding best practices for researchers looking to enter this challenging application area.

Retrospective Submission Format

In order to facilitate responses from as diverse a group as possible, we will accept submissions of multiple forms. Submissions may be in one of the following submission types (all will be submitted in pdf format):

• Abstract of 250-500 words.

• Slide deck with up to 12 slides (and optionally 3 additional supplement slides).

• A standard NeurIPS format poster (36W x 48H inches)

All content must be self-contained  -- e.g. slide decks should contain all context necessary for reviewers to understand the content without a presenter.

Similar to the call for papers, retrospectives should be submitted to the CMT instance linked above.

Retrospective Review Criteria

• Clarity of exposition of the original AI + HADR application

• A critical and holistic assessment of the points of success and points of failure after the original AI + HADR project ended and/or the system was deployed.

• Potential for lessons learned to help guide best practices for future AI + HADR endeavors.

3. Call for Challenges

This year we are additionally announcing a call for challenges: brief abstracts that detail opportunities where the intersection of AI + HADR could be particularly impactful, or is currently not living up to its full potential. The goal for this submission format is to introduce new ideas and diverse viewpoints to the research community’s discourse and encourage networks of researchers who can work together to make a difference on meaningful, but hard, AI + HADR problems. 

Challenge Submission Format

Challenge submissions should be abstracts between 250 and 500 words, not including references (submitted via  Google Form linked here: https://forms.gle/Nj34xpoSpxtkhr6q8.

Challenge Review Criteria

• Clarity and depth of exposition the proposed challenge

• Feasibility of the proposed challenge, with specific action steps or ideal outcomes listed (where applicable).

• Novelty of the proposed challenge and diversity of viewpoints presented in the challenge.

• Potential to to spur new collaborations and insights across AI experts and HADR practitioners.

Accepted Works

Authors of accepted works will be notified on October 23rd, 2021.  Upon notification, we ask that authors of accepted works deanonymize their works, make any final changes, host their work on the web (for instance, on arXiv), and then submit a link to where the work is hosted to the workshop organizers by November 6th, 2021.  The workshop website will then be updated with links to accepted papers.

Note that accepted works will not be formally published. This means that:

• Authors can retain full copyright of their works

• Work contained in accepted papers are not precluded to be published in other research venues

• Submitted papers are allowed to have significant overlap with previously published or currently submitted work (in this case, please indicate overlapping works).

Some or all of accepted works will be selected for oral presentations during the workshop.  Because the workshop is entirely virtual this year, these oral presentations will be presented in video form.  The authors will record a video of themselves giving their presentation and submit it to the program committee by November 1st, 2021.  More specific video submission instructions will be provided in the near future.