AI Can Reduce Prejudice. It Could Also Manipulate Voters.
Also: the myth of “voting against your interests” and who blocks whom more — BCB #153
One of the more effective strategies for reducing prejudice among people is “deep canvassing,” or fostering thoughtful, empathetic conversations that encourage participants to share something of themselves and to consider the personal stories of their interlocutors. But this is very time intensive and doesn’t scale. So recently, a group of researchers tested whether AI could be employed to perform deep canvassing, and found that it can.
The researchers asked over a thousand diverse participants to have a dialogue with an LLM trained to conduct deep canvassing-style, narrative-driven conversation about unauthorized immigrants and support for immigration policy reform. For example, the model prompted participants to articulate their stances on immigration and share the observations and experiences that led them to hold their beliefs. It also shared relevant stories to drive the conversation.
Researchers found that right after the conversations, anti-immigrant prejudice fell and pro-immigrant policy support rose. Those attitude shifts held in a follow-up assessment five weeks later, which happened to be during the final month of the 2024 election cycle.
This demonstrates the persuasive potential of AI. Employing humans to conduct deep canvassing is comparatively time- and energy-intensive, so this could be a way to scale such prejudice-reduction techniques.
Nonetheless, we should expect a fight over the use of this kind of technology. While treating individual immigrants as people rather than stereotypes is hard to oppose, the observed effects on immigration policy may raise concerns about political influence. Many people do not agree with the policy ideas considered “pro-immigrant” in this study (bear in mind items 2 and 3 are phrased in reverse):
1. “The government should provide legal aid to undocumented immigrants who cannot afford an attorney for deportation proceedings.”
2. “Local police should automatically turn undocumented immigrants over to federal immigration officers.”
3. “The federal government should work to identify and deport all undocumented immigrants, including in the workplace.”
4. “The federal government should grant legal status to people brought to the U.S. illegally as children…”
5. “Undocumented immigrants should be able to become citizens after five years of work and tax-paying.”
The AI was provided with “three exemplar narratives illustrating immigrant hardships.” If it was instead provided with negative personal stories about people’s experience with immigrants (draining local government resources, housing shortages, crime, etc.) would the persuasive effects be similarly large in the other direction?
So in the end, this will probably be a fight over which kinds of personal stories are considered representative (e.g. undocumented immigrant crime rates are lower than citizens, but not zero) and whose rights and suffering are considered important. This sort of persuasive technology will never be limited to one side, and each side will say the other side’s use is misleading, unethical, or unfair.
And yet… increasing empathy and understanding of the other seems like it isn’t a bad thing. Can we reliably draw a line between deeper mutual understanding and machine-driven political persuasion?
Do working-class people really “vote against their interests”?
Progressives often argue that working-class Americans “vote against their own interests,” noting that Republican tax policies tend to favor the wealthy and often involve cuts to programs like Medicare and Social Security. But as Justin Vassallo writes in The Liberal Patriot, this view may miss important aspects of what working-class voters value. For all the Democrats’ emphasis on saving these programs, they “are less open about what they would do to restore economic opportunity in distressed regions and strengthen Americans’ pride in where they live.”
As Vassallo explains, twentieth century Democrats sought to build a system that would promote economic prosperity and opportunity for all Americans, modeled on parts of Western Europe. But, he adds,
the vision of a strong welfare state was tied to the expectation that wages, public health, education levels, and other key metrics would be standardized over time, thereby eliminating once-major gaps between regions and demographics. In short, Democrats had an abiding belief in the importance, and feasibility, of continued enablement through decent jobs, schools, housing, and affordable family recreation.
But over time, regional gaps widened instead of narrowing. Coastal cities benefited from globalization and the tech economy, while many post-industrial regions saw job losses, population decline, and emptying main streets. As these trends accelerated, Democrats often struggled to offer clear plans for revitalization—and to many voters in these regions, the party began to feel distant from their day-to-day concerns.
The intensifying reliance on heavily-curated expertise, combined with the party elite’s highly-credentialed cultural milieu, put Democrats in a box. They had become at once too business-friendly and divorced from the big developmentalist ideas that had guided the successful public-private partnerships of the past. Their recurrent emphasis on the social safety net, meanwhile, bespoke a guilty conscience more than a determination to prevent bad outcomes. Mainstream economics, even of the most liberal variety, dictated that the jobs which had provided the foundation for postwar prosperity were obsolete, leaving many well-meaning Democrats at a loss over how to talk about a crisis of development in the world’s most advanced economy.
By contrast, Trump and Republicans repeatedly talk about rural areas, bringing blue collar jobs back to American, and “the forgotten man and woman.” Right now, many people believe Democrats to be the party of the elites—and they’re not entirely wrong.
Who blocks whom more?
Traditional social and political psychology research has often contended that conservatives dislike liberals more than vice versa. But new research offers reason to believe that actually, it’s liberals who dislike their political opponents more than the reverse.
Researchers selected 4,000 Twitter/X users, representing a balance of Blue- and Red-leaning accounts who were then randomly assigned to be followed by one of six accounts. Three of these were crafted to represent Democrats and three were meant to be Republicans. These accounts retweeted corresponding partisan content.
The idea was that participants would immediately see that their new follower was someone they either strongly agreed with or disagreed with politically.
The research team found that both groups were equally likely to follow back accounts they agreed with, but that Republicans were significantly more likely to follow back Democrat accounts than vice versa. They also found that Democrats were more likely than Republicans to block accounts belonging to their political opponents.
Quote of the Week
