Assessing transparency of corporate digital analytics – Scholar Q&A

It’s a near certainty that you have a social media account and/or have made an online purchase in your lifetime. But how certain are you about how those companies store or use your data? Sure, you clicked the “Terms and Conditions” button, but do you know if the company might use your data for digital analytics? To develop artificial intelligence (AI)?

With data harvesting and digital analytics increasing in scope and capability, a team of first-time Page Center scholars is leading a project that will deliver best practices so companies can be transparent about their data use. To do this, scholars Jonathan Obar, Giuseppina D’Agostino and Motunrayo Akinyemi of York University (Toronto, Canada) are assessing the data analytics transparency of 20 companies. The study is a part of the Page Center’s 2023 call for research proposals on digital analytics.

Please share the origin story of your Page Center project.

Obar: At York University, we are increasingly studying AI governance. There are a number of very interesting initiatives here. The Connected Minds Project, recently launched via the largest federal research grant received in York’s history, includes studying the ethical development of artificial intelligence. Professor D'Agostino is one of the leaders of the initiative. There are related initiatives like York University’s Centre for AI & Society. One of the center’s goals is to support collaborations that get engineers, social scientists and legal scholars talking to each other about moving AI governance research and knowledge mobilization forward. When we saw the Page Center call for research on transparency with regards to data analytics, it seemed like a great opportunity to do just that, connected to York’s ongoing AI governance work. This project also builds on my own published research addressing data privacy transparency and meaningful consent online. Some of this work can be viewed at www.biggestlieonline.com, a knowledge mobilization site funded by the Office of the Privacy Commissioner of Canada.

How does your project fit into the research call on digital analytics?

Obar: Transparency and explainability are fundamental to international calls for AI governance. For example, in its Blueprint for an AI Bill of Rights, The White House lists “notice and explanation” as one of five central principles for AI governance. In its description of the principle it notes, “You should know that an automated system is being used and understand how and why it contributes to outcomes that impact you.” This suggests that what a company says about its use of analytics is vital to consumer understanding and can support safe and informed choices online.

With this in mind, our project assesses what companies say about the use of data analytics, and whether explanations align with calls like those from the U.S. government, and similar positions expressed in the EU and Canada. This project emphasizes that companies attempting transparency should provide consumers with information about data collection, management, retention, and use, as well as about algorithms, and how resulting analytics lead to implications, benefits, and risks for the consumer. A thorough data analytics transparency should address each of these components, and more.

Can you tell us more about the legal aspect to this?

Akinyemi: The legal aspect is one of the reasons I joined the team. Self-managing privacy is an ongoing discussion amongst policymakers and practitioners, and is very relevant to current AI governance debates internationally. The law provides users with rights to access, check, and correct data as well as make decisions about how to manage data and how data is shared with social media giants, such as Meta and TikTok. However, studies have shown that privacy policies are either too short or too long, or they don’t explain enough or they're too complex. Based on the academic literature and what we've assessed so far with our project, we've been able to understand the gaps and the complexities identified in privacy policies. While we are still completing our analysis, preliminary results suggest many privacy policies appear to have gaps when it comes to analytics.

Why should I care if a company is transparent about its use of analytics?

Akinyemi: If my data is with a company, I want to know how the data is being managed and what it is being used for. When I sign up for a social media service, for example, and send them photos, I assume those photos will help populate my social media profile. If a social media company intends to analyze my photos using a form of analytics to determine characteristics about me, I should know that before I send the photos. Secondary uses that are not disclosed are a considerable concern because people don’t realize what they are agreeing to when they join digital services. This could provide companies with a lot of flexibility to do a variety of things with the data, some perhaps not in the consumer’s best interest.

Obar: Indeed, transparency is essential if companies are going to pursue an ethical approach to analytics and AI that ensures oversight now and in the future. In terms of the potential for harms, it is important to emphasize that the literature suggests that members of vulnerable and marginalized communities are the most likely to experience harms when it comes to big data and algorithmic discrimination. So equitable approaches to addressing these issues are vital. It could be inaccurate data that’s the problem, but it's not just the data. The algorithms could also be biased. What about the people interpreting the results created by the analytics? They could be biased too.

These concerns are amplified as they spread across companies. Different companies are going to do different things with the data. They will employ different algorithms. They will have different people interpreting the results. Beyond the information provided to consumers, there needs to be strategies for helping people realize oversight protections. Perhaps relying only on text-based forms of transparency is not enough. Companies want to benefit from big data and algorithms, but transparency that is usable is essential if we are going to protect civil rights and civil liberties.

Please talk about the companies you'll be examining. What are you looking for?

Obar: Policymakers have placed notice and transparency at the top of the list of priorities for AI governance. So we've started by going to company websites and reviewing privacy policies. We are conducting a plain language analysis to see how complicated the policies are. Including the plain language assessment, we have 10 transparency criteria we are assessing qualitatively. Essentially, we will review the materials we find to see what companies are saying about the algorithms being used, as well as things like data retention and secondary use.

Akinyemi: We are reviewing social media, e-commerce, and brick-and-mortar companies offering digital services. Any of these organizations might be using AI as we speak, and may intend to benefit from data analytics in the future. A central goal with this project is to assess how transparency efforts are going, with the hope that we can encourage improvements, if necessary.

What are your expectations? What type of results do you think this project will deliver?

Obar: Prior research on data privacy transparency suggests that companies are reluctant to provide considerable detail about technologies that raise privacy concerns. We hope to help advise companies as they consider self-regulatory efforts. As they attempt to be compliant with what policymakers are recommending, how can they be transparent? For example, how should companies explain their algorithms? How is automated decision-making used? Our research over the years was originally guided by the Electronic Frontier Foundation, which is an American civil society group. They first organized a transparency report in 2011 called “Who has your back?”. What the EFF found, and what our research continues to suggest, is that academic research can help guide companies as they try to figure out how to improve transparency efforts. Whether it’s plain language or including details that consumers need to make meaningful decisions, these are the types of things that should be considered by companies when we share our results.