Challenge Statement

The rapid growth and adoption of emerging technologies are reshaping the future of work. This creates unique and complex challenges for people with disabilities. [5] In particular, Artificial Intelligence (AI) is woven into recruitment practices, serving as a gatekeeper for every step of employment. But bias in AI systems can lead to discriminatory practices that disproportionately affect people with disabilities. [6], [7]

The impacts include (but are not limited to) exclusion from job opportunities, algorithms that stereotype (unfairly screening out candidates), inaccessible technologies and lack of representation on teams developing AI tools. [8], [9]

For this reason, MaRS and CIBC are seeking solutions that prevent and/or reduce AI bias and the impacts it has on the recruitment of people with disabilities.

Solutions will need to:

• Address AI bias in recruitment, and
• Be led by or developed in direct partnership with people with disabilities

What do we mean by “AI bias”?

AI bias refers to situations where AI creates unfair outcomes, such as reflecting unjust prejudices or reinforcing existing social inequalities. It occurs when an AI system produces results that are systematically prejudiced due to erroneous assumptions in: the machine learning process, training data, or the design of the algorithms. [10], [11]

AI bias in recruitment refers to the unfair and discriminatory outcomes in hiring processes resulting from algorithms that have been trained on historically prejudiced data, leading to a perpetuation of workplace discrimination, lack of diversity, and missed opportunities for underrepresented candidates. [12]

Solution Streams 

We welcome both technological and non-technological solutions to this challenge and offer two streams of prizing, depending on the stage and nature of the solution. 

Stream A: Non-Technological Idea Generation

This stream is dedicated to cultivating new non-technological solutions that prevent and/or reduce AI bias in recruitment for people with disabilities. New ideas/solutions means that the proposed solution has not been tested or implemented previously.

Examples of non-technological ideas that prevent and/or reduce AI ias in recruitment for people with disabilities include, but are not limited to: 

• Policies guiding AI usage/development;
• Initiatives promoting more responsible AI practices; 
• Methodologies for auditing AI practices among organizations; 
• Educational campaigns promoting inclusive AI;
• Data transparency standards;
• New collaboration/partnership models between experts and practitioners

We encourage anyone with an idea that reduces or prevents AI bias in the recruitment for people with disabilities to apply for this stream. This includes people with disabilities, community-based organizations, students, policymakers, researchers, innovators, advocates, etc.

Stream B: Technological or Developed Non-Technological Solutions 

This stream is dedicated to Technological and non-Technological solutions that have already been initiated to prevent and/or reduce AI bias in recruitment of people with disabilities. Solutions for this stream either have a prototype, Minimum Viable Product (MVP), or are operational in this domain. 

Examples of such solutions include, but are not limited to:

• Specific software applications;
• Innovative modifications (plug-ins or add-ons) to existing AI technologies;
• Tools, inclusive databases, methodologies, policies, or recognized best practices currently effective in preventing and/or reducing AI bias in recruitment for people with disabilities. Crucially, these solutions should have demonstrable evidence of their impact.

Entrants must declare one stream they are entering when submitting their entry. Entrants cannot enter under both Streams.

[5] Bogen, M. (2019, May 6). All the ways hiring algorithms can introduce bias. Harvard Business Review. https://hbr.org/2019/05/all-the-ways-hiring-algorithms-can-introduce-bias.

[6] Caccavale, J. (2022, October 28). Recruitment bias report: How bias affects hiring and how to remove it. Recruitment Software & Hiring Platform. https://www.beapplied.com/post/recruitment-bias-report-how-bias-affects-hiring-and-how-to-remove-it 

[7] Tilmes, N. (2022, April 19). Disability, fairness, and algorithmic bias in AI recruitment – ethics and Information Technology. SpringerLink. https://link.springer.com/article/10.1007/s10676-022-09633-2 

[8] Welker, Y. (2023, May 5). Algorithmic diversity: Mitigating AI bias and disability exclusion. Forbes. https://www.forbes.com/sites/forbestechcouncil/2023/05/09/algorithmic-diversity-mitigating-ai-bias-and-disability-exclusion/?sh=6d9200e6417d 

[9] Nugent, S. E., & Scott-Parker, S. (2022, September 8). Recruitment AI has a disability problem: Anticipating and mitigating unfair automated hiring decisions. SpringerLink. https://link.springer.com/chapter/10.1007/978-3-031-09823-9_6 

[10] Welker, Y. (2023, May 5). Algorithmic diversity: Mitigating AI bias and disability exclusion. Forbes. https://www.forbes.com/sites/forbestechcouncil/2023/05/09/algorithmic-diversity-mitigating-ai-bias-and-disability-exclusion/?sh=6d9200e6417d

[11] Marr, Bernard. (2022, September 30). The problem with biased AIs (and how to make AI better). Forbes. https://www.forbes.com/sites/bernardmarr/2022/09/30/the-problem-with-biased-ais-and-how-to-make-ai-better/?sh=43a7708e4770

[12] Tilmes, N. (2022, April 19). Disability, fairness, and algorithmic bias in AI recruitment – ethics and Information Technology. SpringerLink. https://link.springer.com/article/10.1007/s10676-022-09633-2 d

Target (Pass/Fail)

• Is the solution designed explicitly for people with disabilities?
• Is the solution addressing AI bias in recruitment? 

Innovation (25%)

This criteria evaluates the novelty of the solution:

• What is the solution?
• What’s the value-proposition for people with disabilities? 
• To what extent is the solution different from previous and existing solutions?

Impact (25%)

This criteria evaluates what success looks like:

• What will the solution achieve? (specify outcomes)
• How will the challenge participant measure impact? (the approach or methodology to data collection)

Feasibility (30%)

This criteria will assess the practicality of the solution:

• What is the plan for implementation?
• Is the team capable of delivering this solution?
• How easily can the solution be adopted by other organizations/people/governments?

Inclusion (20%)

This criteria evaluates the approach for considering and including diverse perspectives in the creation of the solution:

• What was the approach to inclusive design when designing, developing and implementing the solution?
• In what ways has the participant taken an intersectional lens in the design, development and implementation of the solution?

The total amount of funding to be awarded is $100,000 CAD 

Stream A will have up to $25,000 CAD in funding

• No more than 5 prizes* awarded in this stream
• Any prize awarded in Stream A will be a minimum of $5,000 CAD

Stream B will have up to $75,000 CAD in funding

• No more than 3 prizes* awarded in this stream
• Any prize awarded in Stream B will be a minimum of $25,000 CAD

Please refer to the “Challenge Overview” tab for more information on each stream and “Challenge Rules” on further details about prizes.

*The actual number and value allocation of prizes for each stream will be determined in the judging phase, based on the strength and number of entries in the relevant Challenge stream.