The Washington PostDemocracy Dies in Darkness

What does informed consent mean when conducting a field experiment?

Sample consent form from an academic study (Source:

Joshua Tucker: The following is a guest post by New York University Ph.D. candidate in political science – and Monkey Cage editorial assistant – Renard Sexton. For readers attending the Midwest Political Science Association annual conference later this week who are interested in discussing this topic further, I invite you to come to a roundtable we have organized on the topic at 4:45 p.m. on Saturday.


As field experiments have become more common in social science research, numerous tricky ethical issues have emerged, none more so than the difficulty of applying the concept of informed consent.

Initially developed with medicine and later psychology research in mind, ethical principles like informed consent have been rolled out to social science scholars as our work has grown to include lab and field experiments. Unfortunately, these principles have not yet been adapted to political science, and thus provide limited guidance as researchers develop their work.

I experienced this disconnect personally while taking several short certification courses on research ethics through the Collaborative Institutional Training Initiative (CITI Program) at the University of Miami, and found that even the political specific training included almost exclusively case studies and guidance from medicine and psychology experiments. The informed consent portions of the training involved reading the relevant law and sample documentation for clinical drugs trials and psychology experiments involving children.

Informed consent, in short, is a process by which a researcher provides the necessary information to a subject about the nature of study such that the subject can competently decide whether to participate or not.

In medicine, informed consent is often relatively straightforward. If a patient is interested in participating in an experimental drug trial, researchers can provide them with the best information they have available about the risks and potential benefits that the drug may have for them. At the end of the day, the individual can decide whether to take part or not, provided they are of sound mind, have not been coerced or deceived, etc.

Political science experiments look different, in several key ways.

First, the treatments that are being tested have effects that spill over from one subject to others – including those who are not even within the experiment. For example, in an experiment that I am conducting in Peru, members of rural mining communities are invited – but not compelled – to participate in an SMS-based communication platform that allows them to exchange messages with the local mining company, with whom communities have had numerous years of conflict, including violence.

Members of a community who choose not to participate in the SMS platform will still experience the effects of the experiment; if the level of conflict decreases through improved information and trust (which is what we predict) or the level of conflict increases, community members who did not personally consent to being in the experiment will be affected.

To deal with this problem, we are approaching entire communities before even the sampling phase of the experiment starts to determine if they are willing to participate. Elected and traditional leaders have been involved in the process, and the kickoff event for the platform involves a public community meeting to which all are invited.

After assembling the universe of communities that are willing to participate in the experiment, we can randomly assign treatment and control. For those who are assigned to control, their addition to the communication platform will be delayed until the end of the data collection period, so that we can test effectiveness, without excluding people that want access.

This is not a solution that works for all experiments; however, it reflects the reality that in experiments where interventions are introduced at a group level, individual informed consent may not be sufficient.

A second difference is that some social science field experiments look more like large-scale public health interventions than laboratory experiments in medicine or psychology, rendering the concept of informed consent particularly hazy. Where a medical trial might do an intensive comparison of two procedures with a sample of 1,200 patients (such as Fishman et al 2003, who compared lung-volume-reduction surgery and continued “normal” medical treatment in patients with severe emphysema) or my Peru experiment would measure intensive data from 1,600 households in the treatment and control groups, large scale voter experiments access exponentially more people.

For example, in their 2008 study of social pressure and voter turnout, Alan Gerber, Don Green and Christopher Larimer sent treatment mailers to more than 80,000 households, and collected data on more than 180,000 households in total. In Facebook’s 2010 study of how social media affect voting, more than 60 million people were included in the study.

In both cases, subjects were neither informed nor did they directly consent to be included in the study. The authors note that in both cases, the interventions administered were very similar to other information that subjects were already receiving, and that at a local level they could choose not to participate. In the first example, voters could choose not to read the mailers sent to them, and in the second case, voters could choose not to be on the Facebook Web site.

Thinking in more practical terms, with sample sizes in the hundreds of thousands or millions, getting expressed informed consent from each subject becomes a massive task that would swamp the resources of even the most robust research budget. With fairly innocuous treatments and huge numbers of participants, many argue, is the traditional informed consent form even relevant?

October’s controversial field experiment in Montana by Stanford and Dartmouth researchers brought these informed consent questions to a head. The experiment, seeking to learn about the effect of partisan information in nominally nonpartisan judicial races, drew the ire of state officials for improperly affixing the state’s official seal to mailers.

More pertinent to the discussion here, even if some Montana voters had consented to being part of the experiment, those who did not would still have to live with the results of the intervention. Local bloggers were particularly incensed that a center-left judicial candidate was being linked to President Obama (who had a nearly 60 percent disapproval rate in the state at the time); something that they suggested might render him unelectable. The mailers went to 100,000 registered voters, which, in addition to making it potentially influential on electoral outcomes in Montana, made it completely infeasible to get traditional informed consent from all participants.

The informed consent concerns in social science field experiments I mention here are obviously just two of a bevy of critiques that have been leveled against field experiments in political science and economics. The effort at a solution for the first problem that we are implementing in Peru is just one of many potential approaches to getting community informed consent.

Indeed, this has been a topic of much discussion and debate within political science, and very smart people across the discipline are endeavoring mightily to discern the best way forward. This weekend at the Midwest Political Science Association’s annual meeting, the editors of the Journal of Experimental Political Science are hosting a roundtable discussion on “Informed Consent in Field Experiments;” Don Green of Columbia University is chairing a more general roundtable on “Ethics of Field Experiments.” Similar discussions are being planned for the American Political Science Association meeting in San Francisco in the summer. For political colleagues that will be at either meeting, we invite your comments, critiques and participation.

The goal of these roundtables is to generate new ideas, aggregate experiences and eventually develop disciplinary standards that will take social science, especially political science, forward from the traditional models of informed consent that come from medicine and psychology to standard practices that actually fit the work we do.