Alphabet Inc – Algorithm Disclosure (2023)

Outcome: 17% vote in favor of the proposal

Whereas legislators, regulators, academics, and civil society increasingly require information to help understand how algorithmic systems can lead to discriminatory and other harmful outcomes in education, labor, medicine, criminal justice, and online platforms.[1]

In 2022 the White House published a Blueprint for an AI Bill of Rights including a call for “algorithmic impact assessments” from independent evaluators to look for discrimination.[2]

In 2021: (1) bipartisan lawmakers introduced the Filter Bubble Transparency Act, which would require companies to provide a version of their products which uses an “input-transparent” algorithm; (2) the Social Media Disclosure and Transparency of Advertisements Act was introduced in Congress and would force disclosure regarding online targeted advertisements; and (3) Washington, D.C. Attorney General Karl Racine introduced the Stop Discrimination by Algorithms Act, which would require companies to audit algorithms for discriminatory impact.[3]

General artificial intelligence bills or resolutions were introduced in at least 17 U.S. states in 2022 and enacted in four.[4]

In the EU, the European Commission is working on an artificial intelligence regulation to address risks associated with uses of AI and to build trustworthy artificial intelligence.[5]

In 2021, an investigation by The Markup found that Google Ads “blocks advertisers from using 83.9 percent of social and racial justice terms”. White supremacist and anti-Muslim ideologies have appeared on YouTube and can lead to offline violence; for example, the New Zealand Royal Commission found that content on YouTube radicalized the man who massacred 51 people at Christchurch mosques in 2019.[6]

In 2020, Google fired Timnit Gebru, co-lead of Google’s AI ethics team, after she conducted research that found Google’s technology could perpetuate racism and sexism.[7]

Promoting fairness, accountability, and transparency in artificial intelligence is central to its utility and safety to society. The Open Technology Institute has recommended a set of algorithmic disclosures for tech companies. Deloitte has said algorithmic risk management “requires continuous monitoring of algorithms”. The Mozilla Foundation and researchers at New York University have put forward recommendations and technical standards for algorithm and ad transparency.[8]

Shareholders believe that improved disclosure will help in building and maintaining users and investors’ trust, that will ultimately drive long-term, sustainable value creation.

Resolved, shareholders request Alphabet go above and beyond its existing disclosures and provide more quantitative and qualitative information on its algorithmic systems. Exact disclosures are within management’s discretion, but suggestions include, how Alphabet uses algorithmic systems to target and deliver ads, error rates, and the impact these systems had on user speech and experiences. Management also has the discretion to consider using the recommendations and technical standards for algorithm and ad transparency put forward by the Mozilla Foundation and researchers at New York University.

[1] https://d1y8sb8igg2f8e.cloudfront.net/documents/Cracking_Open_the_Black_Box.pdf

[2] https://www.whitehouse.gov/ostp/ai-bill-of-rights/

[3] https://finance.yahoo.com/news/bipartisan-bill-seeks-curb-recommendation-225203490.html; https://trahan.house.gov/news/documentsingle.aspx?DocumentID=2112; https://oag.dc.gov/release/ag-racine-introduces-legislation-stop

[4] https://www.ncsl.org/research/telecommunications-and-information-technology/2020-legislation-related-to-artificial-intelligence.aspx#2022

[5] https://digital-strategy.ec.europa.eu/en/policies/european-approach-artificial-intelligence

[6] https://themarkup.org/google-the-giant/2021/04/09/how-we-discovered-googles-social-justice-blocklist-for-youtube-ad-placements; https://www.cnbc.com/2020/12/08/youtube-radicalized-christchurch-shooter-new-zealand-report-finds.html

[7] https://www.wired.com/story/google-timnit-gebru-ai-what-really-happened/

[8] https://www.newamerica.org/oti/reports/why-am-i-seeing-this/promoting-fairness-accountability-and-transparency-around-algorithmic-recommendation-practices/; https://www2.deloitte.com/content/dam/Deloitte/us/Documents/risk/us-risk-algorithmic-machine-learning-risk-management.pdf; https://blog.mozilla.org/en/mozilla/facebook-and-google-this-is-what-an-effective-ad-archive-api-looks-like/; https://foundation.mozilla.org/en/campaigns/mandating-tools-to-scrutinize-social-media-companies/; and https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3898214