The campaign for Republican presidential candidate Ted Cruz is working with a British data company called Cambridge Analytica to develop behavior-based models of American voters.

Created in 2013, the so-called “psychographic” firm collects up to 5,000 data points on every U.S. voter and uses the results of a survey of up to 50,000 people per month to predict their personality type.

When a Cruz campaign worker knocks on a voter’s door, the campaigner knows who the voter is, what issues he cares about, and how to persuade him into voting for the Texas senator.

Is this so-called “voter microtargeting”  the new wave of political campaigning? We asked Nick Beauchamp, an assistant professor in the Department of Political Science and a core faculty member in the NULab for Texts, Maps and Networks who studies how political opinions form and change as a result of discussion, deliberation, and argument.

One microtargeting expert told NPR that big-picture data and computational modeling might not be able to capture the essence of an individual personality, saying that “persuadability is finicky.” In your view, just how much can a voter’s political opinion be changed by a conversation with a campaign worker armed with a script tailored to the voter’s pet issues?

Persuasion is indeed finicky. In some ways it is easier than turning out new voters, since all you have to do is nudge which lever they pull, rather than getting them out of the house and across town to make a hard set of decisions with little direct payoff. But in other ways, it is much harder. Since we all generally agree that voting is a civic duty, boosting turnout can mainly be a job of finding one or a few messages that appeal to everyone. But vote choice is inherently oppositional: What is a good message for a supporter of one candidate can have the opposite effect in an opponent. For this reason, the appeal of microtargeting or tailored scripts becomes much greater in persuasion, since it can (a) boost appeal based on a message that hits on exactly the issues most salient to the listener, and (b) avoid the backfiring effects of hitting the wrong person with the wrong message. But although we all believe this can be done—there are persuadable people, and some do make up their minds based on what they see and hear during a campaign—landing the right message with the right person is incredibly hard; inherent in this is the idea that if you get it wrong, your message may have no effect or even a negative effect, compared to a more general message that appeals (more weakly) to everyone.

Both primaries currently illustrate both sides of this. On the one hand, we have Trump and Sanders, who have broad-based, mass-media messages that appeal to many people equally; on the other hand, we have Clinton and Cruz, who use more elaborately targeted and tailored messages, some of which work very well and some of which fail or backfire. Microtargeting is high risk and high reward, but it is unclear, even after decades of work and in the era of Big Data, whether we have yet figured out how to do it sufficiently well to always outperform the benefits of a more generically appealing message.

According to the Center for Democracy and Technology’s chief technologist, “If people understood that this amount of fine-grained, sensitive data was being used by political campaigns, they would likely feel betrayed.” From your perspective, is voter microtargeting the quintessential form of political persuasion or a dirty trick played on the unwitting public?

What’s interesting about the public anxiety about these things is how ubiquitous these sorts of practices already are in product marketing. Many of the political microtargeting databases were built on top of—or greatly augmented with—existing market-research databases, which have been used to microtarget ads of various sorts for decades. Political persuasion may be more alarming, however, because it pertains to shared, foundational values that could have broad societal effects, rather than just manipulating our snack preferences.

On the other hand, unlike marketing, there are actually good forms of political persuasion. For instance, a campaign that provides valuable and true information, even if it’s dedicated to electing a candidate, can be informing voters in a way that allows them to make better decisions that better reflect their fundamental beliefs. The fear with microtargeting, however, is that it is engaged less in this better, more deliberative persuasion, and more in manipulative emotional effects that nudge people without them understanding the process or learning anything from it.

“Why Big Data matters.” “How Big Data can help solve the world’s woes.” “The importance of Big Data is growing exponentially.” These are just a few of the countless “Big Data” headlines that have been emblazoned on the covers of newspapers and magazines the world over. How, in your view, is Big Data transforming politics, and what will its role be in elections five, 10, even 15 years from now?

Part of the worry about Big Data from the public’s point of view is that both the data and the methods can be obscure and opaque to outsiders, make the same kinds of campaign manipulations that have been going on for decades now appear more threatening and powerful when performed by the terabyte and gigaflop. That’s also part of its appeal and reputation—this idea that somehow enough data will qualitatively change the difficult tasks of modeling personality and behavior. And there is some truth to both sides: Big Data really does allow us to explore issues of opinion and persuasion in ways heretofore impossible, by combining hundreds of variables over millions of people or examining the gigabytes of raw text we generate online every day.

I think the result now is that we understand opinion and personality in much richer detail than before as well as how various aspects of our public and private personae interact, such as voting, buying, work, travel, dating, socializing, and personality. What hasn’t been fully modeled, I think, is how those characteristics affect behavior, and how to intervene and change that behavior, despite many grand promises from numerous political analytics firms. It’s one thing to measure hundreds of variables or run dozens of experiments; it’s another to synthesize that data into general theories that extend beyond a single set of candidates or campaigns. But I believe rapid progress is being made, with the sort of ethical problems that people unnecessarily fear now potentially becoming more of an issue in the next decade. Just as ethical concerns have changed how we collect and share Big Data, those concerns may also in the future shape which sorts of microtargeting, persuasion, and tactics are acceptable in a free and open society in which more and more of us have access to powerful computational persuasive methods—perhaps as $1 apps on our eye-implanted “phones.”