June 14, 2016

Research review at Facebook

By: Molly Cohn Jackman

Today, we’re sharing more information about our research review program. This process reflects input from many subject matter experts from outside the company and we plan to continue to find ways to improve our own process and gather feedback over time. We hope that publishing our current process provides a window into how we incorporate research ethics into our work and that our experience so far can inform other companies and organizations as they develop research review frameworks.

Research is one of the most important tools we have to give people a good experience on Facebook. It helps us understand what’s working and what should be improved. Most research at Facebook is focused on making sure we build products that are responsive to the needs of our community — not just based on our instinct of what our community might need.

We also think research, when conducted in the open, can be important in understanding some of the world’s most challenging problems — from increasing internet access in the developing world to building servers that can connect billions of people to each other. Facebook researchers participate in conferences, publish in peer-reviewed journals, and partner with experts from academic institutions around the world. Innovation occurs most quickly when large and diverse sets of researchers build on each other’s work, and we are grateful for the feedback we receive as members of the broader research community.

Just as we benefit from feedback on our research, we have also benefited from feedback on our research program. Over the last two years, we’ve strengthened our review program to reflect input from academic experts, research ethicists, policymakers, and the people on Facebook. We first outlined our approach to research review in October 2014 and have further improved our program, which you can read more about in Washington & Lee Law Review.

Our program begins with training. All new employees learn about Facebook’s policies around data access and privacy. Anyone who conducts research also attends a bootcamp where they learn specifically about our research review program, why it matters, and how it works.

Before initiating research, an employee must propose the project to their Facebook research lead — a substantive expert in the relevant field who is trained in research ethics. All research proposed at Facebook is subject to this review.

Most of the research we conduct involves standard product improvements — for example, evaluating the effect of changing the size or color of a button. This type of research is common across industries, and it is important for making educated decisions about how to improve people’s experience with our products. However, it rarely raises ethical considerations. In these cases, research leads can expedite the approval of proposed research.

But as Facebook has grown, so has our research agenda, and we’ve learned that serving a global population requires studying complex, challenging questions about people and cultures. For example, our compassion research team studies how Facebook can help people during life’s most difficult moments, so that we can develop tools that address issues like breakups and suicide prevention. Our data scientists have studied how people support friends during hard times. This research helped inform the creation of Reactions, since we found that people needed more ways to respond to negative posts on Facebook.

When proposed work is focused on studying particular groups or populations (such as people of a certain age) or topics that may be considered deeply personal (e.g., how online tools can serve people during life’s hardest moments), the research lead submits it to an extended review group. This group is composed of five experts, with experience across substantive research areas, research ethics, law, and policy. The research lead can send any proposal — regardless of scope or topic — to this group. One of our company values is openness, and we tend to err on the side of over-inclusion. When in doubt, we like to have more eyes on a research proposal than fewer.

The extended review group considers a research proposal’s potential benefits, such as improvements for people on Facebook or contributions to general knowledge. It also identifies any potential costs and risks, such as data privacy or security issues that have not already been reviewed through Facebook’s privacy program, or other issues such as impacts on vulnerable populations. Additional experts — from Facebook and beyond — are often asked to advise on specific proposals.

Research only begins when the full group agrees that the benefits — to Facebook, our community, and our society — are clear and present, and that potential downsides have been addressed.

We have designed our program to be flexible, so we can continue to learn and adapt based on the needs of our community and the feedback we receive. We hope our work will help other companies establish or improve their own programs, so that more organizations contribute to the world’s understanding of important issues.