The Theory

Mounting evidence from a variety of fields—psychology, political science, communications–suggests that facts, alone, rarely suffice to change minds in contested areas. Rather, this research suggests that different “frames” or “narratives” can make contested facts appear more or less threatening, depending upon the context in which those facts are embedded. What seems to matter most for persuasion, then, is whether particular facts are framed in such a way as to support, or challenge, one’s personal and political values.

Consider one example, straight from the social science research on this subject. Yale’s Dan Kahan and his colleagues have shown that those with “hierarchical-individualist” values—conservative, supportive of private industry—are more likely to accept the science of global warming if that science is framed to support an advancement of nuclear power, than if it is framed to support EPA regulatory action to curb fossil fuel emissions.

Such findings suggest that it might be possible to counter rampant misinformation by selectively framing contested facts, and targeting particular audiences where resistance levels are known to be high–because these facts are currently perceived to conflict with their values. But how to reach those audiences, and know their values in advance?

Enter our hypothetical Facebook app—“Value Ads”–designed for #Truthicon. With such an app, we could target individual Facebook users who have a known set of values, and deliver them non-threateningly framed factual information through Facebook ads. For so-called “hierarchical individualists,” for example, we could frame the facts about global warming in the context of advancing nuclear power—an argument they might find more convincing.

Measuring Values Online

Value surveys and political data processing are well established techniques for personalizing reader experience. These have been made famous in the US by OK Cupid and Hunch. In the UK, Preloaded and Channel Four have developed “The End,” a beautiful videogame which invites teenagers to reflect on the meaning of mortality. As the game unfolds, players are exposed to information about famous figures who share their beliefs. They are also offered opportunities to explore different perspectives through role play.

Yahoo Research has already experimented with categorizing the most politically polarized searches, with the possibility of offering fact checks alongside search results. (Have they ever considered offering politically-personalized targeted ads?) Finally, A Thin Line, developed by MIT Media Lab researchers and MTV, helps inform young people when they’re at risk of going over the line with online communications.

At Truthicon, our team didn’t have access to the vast datasets of Google, Yahoo, or Facebook. So we needed another way to (a) develop a dataset of political values and (b) correlate those values to the factors actually used to personalize online ads. If we can correlate political values to certain kinds of Facebook “likes” or search terms, for instance, then we can develop Google AdWords and Facebook Ad strategies which are likely to reach people holding these political values. Our ads can then include informational material which is framed in terms that someone with these values is more likely to consider.

The system starts with data collection. Our team decided to create a political quiz for social networks, using the basic twelve question scale that Kahan and his colleagues ask to determine whether one is “hierarchical” in outlook or “egalitarian,” and whether one is “individualistic” in outlook or “communitarian.” For the questions themselves, see the supplementary materials in the research paper downloadable here. It would certainly be possible to use different values measurements or schemes for a different quiz, but we chose the Kahan et al framework because of its simplicity and clarity, and because it has been shown to lead to successful framing interventions in experiments.

When people take the quiz, it rates their social and political values and lets them compare their results with those of their friends. Social networks already have many of these kinds of quizzes. In an election year, many people will genuinely be interested in this form of self discovery. In addition to collecting information on people’s political values, our app could request information on the books, concerts, bands, and other things they like. To keep the process transparent, we could share that information back to respondents, showing them what books, music, or websites are most polarizing, and which ones bridge ideological divides. Obviously, the respondents would be offered the opportunity to decline to share this information with the app.

Correlating Values and “Likes”

Let’s assume our Facebook survey game is enjoyed by a large, diverse sample of the American public. What happens next is the hard part. At this point, most systems would try to deliver personalized material to users of our app. That’s certainly one approach, but at Truthicon, we imagined reaching a much broader audience.

Assuming the software finds good correlations between political values and particular “likes” on Facebook, we could use it to plan AdWords and Facebook Ad campaigns that reach much broader audiences. For example, if egalitarians prefer strawberry jelly over grape, then a media campaign aimed at informing egalitarians could be directed towards everyone on Facebook who likes strawberry jelly. What if jelly preferences are less important than taco crunchiness? The “Value Ads” software could propose the top 100 preferences which are most likely to reach the greatest number of egalitarians.

We would need to be sensitive to privacy concerns. Although there are many bad players in online advertising, social networks also allow app developers to show responsibility and transparency. In this respect, a focus on the public interest aspect of this endeavor will be important. In projects like the game “The End,” for instance, any data collected is mirrored back to the player as part of a process of personal discovery. We could do the same. And again, we would ask to collect the data in the first place.

The websites to which our Values Ads point also need to be open, and in the public interest. It’s possible to create spaces on the web where people have genuine dialogue and where good information prevails. Values Ads could direct readers to those communities. Patheos, for example, is an amazing community for multi-faith religion dialogue online. PolicyMic is trying to cultivate a similar non-partis,an politics debate community. Alternatively, sites like Change.org succeed because each piece of content is transparent about its political aims.

Values Ads are a non-trivial technical challenge, but we think the greatest hurdle might be to develop a strong, non-partisan web brand that can offer facts–personalized to divergent values. If we can do that, the facts might just become easier for anyone to swallow.