A New Approach to Evaluating Social Media Ethics

The concept of ESG (Environmental, Social, and Governance) ratings for companies, including assessments of their carbon footprint, is now widely recognized. Recently, a U.K.-based firm has pioneered a method for evaluating the ethical standing of social media platforms.

Introducing EthicsGrade

EthicsGrade functions as an ESG ratings agency with a specific focus on AI governance. Led by Charles Radclyffe, previously the head of AI at Fidelity, the organization employs AI-powered models to deliver a comprehensive ESG assessment. This is achieved through the utilization of natural language processing for automated analysis of extensive datasets, including tracking contentious issues and official statements.

Radclyffe’s motivation stemmed from observing instances of “green-washing” within environmental stocks. He identified a critical gap in the evaluation of AI governance within social media companies, recognizing the substantial risk to investors following events like the Cambridge Analytica manipulation of Facebook during the U.S. election and the Brexit referendum in the U.K.

EthicsGrade Social Media Industry Scorecard

The intention behind these ratings is to assist companies in identifying areas for improvement. Furthermore, asset managers can leverage these insights to pinpoint potential AI-related risks within their investments.

The Genesis of EthicsGrade

In an interview with TechCrunch, Radclyffe explained that during his time at Fidelity, he became the primary resource for colleagues seeking to understand the risks associated with technology firms. Repeated inquiries regarding questionable facial recognition companies and social media platforms highlighted a significant lack of available data, relying instead on anecdotal evidence.

Following his departure from Fidelity, Radclyffe established EthicsGrade to encompass not only traditional ESGs but also AI ethics for algorithm-driven platforms.

He elaborated: “We have developed a model to analyze technology governance, covering 20 different industries. Much of our published work thus far has focused on non-tech companies, as these risks are prevalent across various sectors. However, we are preparing to launch our data specifically related to technology, beginning with social media, in the coming weeks.”

A Parallel to ESG Ratings

The approach taken by EthicsGrade closely mirrors the established practices within the ESG space.

Radclyffe stated: “Our aim is to answer questions such as: how does TikTok compare to Twitter, or WeChat to WhatsApp? We’ve observed that regulations like GDPR have significantly improved standards regarding data privacy and governance. However, other areas, such as ethical risk and a company’s approach to public policy, involve technical questions of risk management.”

Evaluating Algorithms Themselves?

The core function of EthicsGrade involves assessing algorithms. The company confirms that it is developing its own AI through NLP to automate the process currently reliant on human analysts, mirroring the evolution seen in the environmental assessment field with firms like “sustainalytics.”

EthicsGrade’s rating methodology centers on evaluating “the degree to which organizations implement transparent and democratic principles, guarantee informed consent and risk management procedures, and foster an environment conducive to error correction and continuous improvement.” This evaluation is based entirely on publicly accessible information – policies, websites, lobbying efforts, and more. Essentially, they assess the governance of the AI, focusing on the safeguards in place to ensure ethical outcomes and responsible input management, rather than the algorithms themselves.

Targeting Asset Owners and Managers

“Our primary objective is to reach asset owners and asset managers,” Radclyffe emphasized. “For instance, Twitter is 29% owned by five organizations: Vanguard, Morgan Stanley, Blackrock, State Street, and ClearBridge. The ownership structures of Facebook and Microsoft are similar, with Fidelity, Vanguard, and BlackRock holding significant stakes. Therefore, we need to gain the support of a limited number of key players by demonstrating that the questions journalists have been raising for years are relevant to their portfolios.”

When questioned about analyzing the content of posts, such as Tweets, Radclyffe clarified: “We do not analyze content. Our focus is on how they govern their technology and where we can find evidence of that governance. We share our rating and assessment with each company, emphasizing that it is based on publicly available data. We then invite them to complete a survey, which helps us validate our findings. Microsoft is currently the only company to have completed the survey.”

The ideal outcome is for companies to “verify the information, confirming that they have established processes to ensure effective management and prevent algorithmic bias.”

The Future of Algorithm Risk Assessment

As algorithms increasingly shape our world, the viability of rating them for risk, particularly among asset managers, remains to be seen.