Despite a rapid expansion of machine learning (ML) across fields and industries, it is not seen as understandable by the general populous. A 2017 study by The Royal Society interviewed members of the public in the UK, finding that a majority of participants knew “little to nothing” about machine learning.1 While many of the participants were aware of technologies that use ML, very few were aware of how the technology worked, “even at a broad conceptual level.” Another study found that even amongst UX designers working on projects that involved ML, a lack of understanding was common. One participant referred to ML as “black magic,” stating that “designers don’t understand the constraints of the technology and how to employ it appropriately”.2
As ML technologies appear in more everyday contexts and make increasingly consequential decisions in our lives, this lack of understanding is troubling. Regulation around these technologies is nascent, and as policymakers think about reasonable legal structures, the public should be empowered to engage in these discussions.
Widespread public engagement can also help disrupt a harmful power dynamic between the developers of automated systems and the people that these systems affect. The people and companies building ML technologies are a homogenous population that are overwhelmingly wealthy, white, and male. In 2018, only 18% of first authors at 21 ML conferences were women,3 and only 2.5% of Google’s workforce was black.4 A recent report by the AI Now Institute details the way in which this lack of workplace diversity is fundamentally tied to gender- and race-based discrimination in systems themselves.5 The general lack of understanding of how these systems work only exacerbates the power differential. If individuals and communities are able to critically understand the impact and limitations of ML, they can hold developers and companies accountable, and perhaps feel empowered to build grassroots technologies themselves.
Understanding begins with education, but many popular and currently available materials for learning ML are too technical to be accessible to a general audience, too broad to be useful, or simply wrong. As ML researchers, we hope to fulfill this need for good learning materials that can be useful for a broad audience.
Check out our paper pre-print for more info!
1. Ipsos, M. O. R. I. "Public views of machine learning." Royal Society (2017).
2. Dove, Graham, et al. "UX design innovation: Challenges for working with machine learning as a design material." Proceedings of the 2017 chi conference on human factors in computing systems. ACM, 2017.
3. Element AI. (2019). Global AI Talent Report 2019. Retrieved from https://jfgagne.ai/talent-2019/.
4. Google. (2018). Google Diversity Annual Report 2018. Retrieved from https://static.googleusercontent.com/media/diversity.google/en//static/pdf/ Google_Diversity_annual_report_2018.pdf
5. West, S.M., Whittaker, M. and Crawford, K. (2019). Discriminating Systems: Gender, Race and Power in AI. AI Now Institute. Retrieved from https://ainowinstitute.org/ discriminatingsystems.html.
As ML technologies appear in more everyday contexts and make increasingly consequential decisions in our lives, this lack of understanding is troubling. Regulation around these technologies is nascent, and as policymakers think about reasonable legal structures, the public should be empowered to engage in these discussions.
Widespread public engagement can also help disrupt a harmful power dynamic between the developers of automated systems and the people that these systems affect. The people and companies building ML technologies are a homogenous population that are overwhelmingly wealthy, white, and male. In 2018, only 18% of first authors at 21 ML conferences were women,3 and only 2.5% of Google’s workforce was black.4 A recent report by the AI Now Institute details the way in which this lack of workplace diversity is fundamentally tied to gender- and race-based discrimination in systems themselves.5 The general lack of understanding of how these systems work only exacerbates the power differential. If individuals and communities are able to critically understand the impact and limitations of ML, they can hold developers and companies accountable, and perhaps feel empowered to build grassroots technologies themselves.
Understanding begins with education, but many popular and currently available materials for learning ML are too technical to be accessible to a general audience, too broad to be useful, or simply wrong. As ML researchers, we hope to fulfill this need for good learning materials that can be useful for a broad audience.
Check out our paper pre-print for more info!
1. Ipsos, M. O. R. I. "Public views of machine learning." Royal Society (2017).
2. Dove, Graham, et al. "UX design innovation: Challenges for working with machine learning as a design material." Proceedings of the 2017 chi conference on human factors in computing systems. ACM, 2017.
3. Element AI. (2019). Global AI Talent Report 2019. Retrieved from https://jfgagne.ai/talent-2019/.
4. Google. (2018). Google Diversity Annual Report 2018. Retrieved from https://static.googleusercontent.com/media/diversity.google/en//static/pdf/ Google_Diversity_annual_report_2018.pdf
5. West, S.M., Whittaker, M. and Crawford, K. (2019). Discriminating Systems: Gender, Race and Power in AI. AI Now Institute. Retrieved from https://ainowinstitute.org/ discriminatingsystems.html.