Researchers pull back the curtain on online conversations
Last fall, thousands of Central American migrants were moving northward through Mexico. Allegations that the group included “criminals and unknown Middle Easterners” spread quickly online. Most of the rumors about the caravan were driven by social media bots, with an estimated 60% of the Twitter chatter from bots.
In general, bots (short for robots) are snippets of software designed to carry out online tasks. Sometimes they’re helpful, speeding up routine tasks like invoice processing and finding cheap flights. Sometimes they meddle, amplifying and spreading divisive messages online.
Faculty and students in the College of Letters and Science are developing sophisticated analytical tools to study the effects of bots. Their findings are opening new avenues for combatting fake news and show how bots can be used in positive ways — from helping people adopt healthy habits to envisioning a digital democracy.
From yellow journalism in the 1890s to the Nazi propaganda machine, the spread of misinformation is nothing new in human society. But bots bring new challenges.
For example, bots can give the false impression that a photo or story is highly popular and endorsed by many, whether or not the information is real. This exploits a human bias: people judge information more favorably if their social circle supports it. “People think messages with more views or likes are more credible,” Zhang said.
Human biases can also be leveraged to help us behave in more health-conscious ways. Seeing one’s co-workers, friends, and family take an exercise class or lose weight encourages others to do the same. In research with collaborators at the University of Pennsylvania, Zhang showed peer influence also works within online social networks, spurring young adults to exercise more.
Zhang is currently looking at ways to co-opt our biases to combat anti-vaccine messages. In one recent study, she tested a two-step strategy that laid out anti-vaccine messages and then refuted them point by point. “The results were concerning but also validating,” Zhang said.
It turns out that exposure to misinformation, even when accompanied by detailed fact-checking, diminished participants’ attitudes about the benefits of vaccines. The study also revealed the culprit. The fake stories about harm from vaccines made people feel angry, rather than fearful or unsafe. The findings correspond to seminal research in communication, Zhang said: “The reason misinformation spreads so fast is it contains an emotional component.” A possible solution is providing pro-vaccine messages that evoke an emotional response, an approach Zhang plans to test.
When bots stoke anger and fear by manipulating news and social media, it’s not just for one side. During the debate around a 2015 California bill that eliminated the personal-belief exemption to mandatory vaccination rules, bot accounts tweeted both pro- and anti-vaccine messages.
Research by Associate Professor Amber Boydstun, Department of Political Science, and Professor Alison Ledgerwood, Department of Psychology, helps explain why antagonizing all sides is a successful strategy — people pay more attention to negative information online.
“Social media is good at magnifying outrage and anger because that tends to be what we click on,” Ledgerwood said.
In collaborative research probing positive and negative messages about political candidates, Boydstun, Ledgerwood and alumna Jehan Sparks (Ph.D., psychology, ’18) found people’s biases form quickly and are resistant to change. Once study participants had negative views about a candidate — even a member of their own political party — it was hard to change their minds. “It is disconcerting to realize just how easily and deeply misinformation can lodge in people’s minds, often with a partisan bias,” they concluded.
Ledgerwood is interested in how to get positive information to stick in people’s brains. “It’s an open question,” she said. Positive political campaigning does work in some circumstances, she adds. “Obama was elected on hope and change, and that positive message was able to get people excited,” Ledgerwood said. “Enthusiasm is what drives voter turnout, so there’s a possibility of motivating political engagement through getting people excited instead of getting people angry.”
Catching democracy up to technology
Recognizing the malicious intent of election interference by bots, there are efforts by lawmakers to reign them in at the state and federal levels. In 2018, Sen. Dianne Feinstein (D-California) introduced the Bot Disclosure and Accountability Act, which would prohibit candidates and political parties from using bots for campaigning. It would also limit PACs, corporations, and unions from using bots for political advertising.
While the damage wrought by bots is often at the forefront of research, Professor Martin Hilbert, Department of Communication, believes bots could help build a more transparent and equitable democracy. “Democracy was created when information traveled by horse. Nowadays information transfers at the speed of light,” he said. “We could build technology that fosters democracy.”
Hilbert, a former United Nations officer, and a team of five UC Davis students (undergraduate and graduate) are developing a big data online observatory for the U.N. Secretariat in Latin America and the Caribbean. The project uses machine learning to sift through vast amounts of public online data to inform international development policies. Machine learning means having algorithms learn in a similar way to humans — observing the world through data and identifying patterns. Hilbert and his students are part of a larger trend of public agencies turning toward big data to understand the wants and needs of a population. The results can provide a broad picture, but the techniques also require careful oversight.
When bots and algorithms get unfettered access to humanity’s unfiltered data, they will mimic human prejudices about gender, age, ethnicity, religion, and more. An experimental Microsoft chatbot called Tay lasted only 16 hours on Twitter before it starting spouting racist tweets.
Hilbert has some ideas for avoiding this problem. Algorithm-driven decisions should be transparent and open to challenge. “If someone disagrees with an algorithm, just like when someone disagrees with a court ruling, why not appeal?” he said. And governments should invest in programmers, rather than relying on private companies, he said. “It’s just that nobody has put the lucas [money] on the table to develop algorithms to promote democracy.”
Watching the watchers
No technology, even bots, is inherently good or bad. But neither is it neutral. As the use of algorithm-driven technologies grows, the potential for abuse increases as well. For instance, Facebook users know the eerie feeling of seeing hyper-targeted advertisements that seem related to their offline conversations.
The age of social media has rendered anonymity almost impossible except for the privileged, said Professor Kriss Ravetto-Biagioli, professor of cinema and digital media and science and technology studies. “The awareness of ubiquitous surveillance modifies our behavior, our speech, and possibly even our willingness to be politically active,” she said. Ravetto-Biagioli is examining the pervasive loss of personal privacy for her latest book project, Resisting Identity.
Both Ravetto-Biagioli and Hilbert agree that social media users should have greater control over their personal data, and that public policies are needed to regulate misinformation on social media platforms. “Technology is not something that falls from the sky,” Hilbert said. “We can create the future we want.”