018 - Elections & Democracy
A path to augmented citizenry?
Editors Note: This blog article is a distillation of an in-person event held in San Francisco on 2024-07-25 facilitated by
. Quotes are paraphrased from the original conversation.
In an era of increasing political polarization and widespread disillusionment with democratic institutions, could artificial intelligence offer a path to reinvigorating civic engagement? This question was at the heart of a recent AI salon discussion, where technologists, policy experts, and civic-minded individuals grappled with the potential impacts of AI on democracy and civic participation.
👉 To jump directly to a list of takeaways and open questions, click here.
Civic Empowerment
Our conversation began by considering citizenship itself, and how more people could be empowered to become part of the influential body that changes our governments. We immediately established that while voting and elections are critical actions for the citizenry, they are only one component, perhaps not even comprising (as one participant put it) 50% of our civic power. Thus a founding question was how can we create a broader empowered citizenry in society - a broad group of people deeply engaged in civic matters who understand how to transform their own values and visions into meaningful action at a government level.
Ideally, AI could support the expansion of this empowered citizenry by processing vast amounts of information, providing personalized civic education, simplifying complex topics, and offering multi-perspective analyses. The goal is to use AI as a tool to lower barriers to civic engagement, potentially creating a more informed and empowered citizenry, thus broadening the empowered beyond a privileged few.
Tool vs replacement?
However, a central tension that emerged from the conversation was whether AI represents a tool for enhancing human civic engagement or a potential replacement for traditional forms of participation. On one side of the debate, were those arguing for the possibility of a broader empowered citizenship who can actually observe and oversee the workings of government. As one participant noted:
"AI never gets tired of paying attention to what's going on in government. It can read all the laws, attend every meeting, read every news article, and synthesize those things."
The idea that government becomes more transparent and accountable through this kind of “citizen surveillance” was enticing. Often, AI surveillance is brought up as a privacy risk, enabling more government control - certainly a worry especially in more authoritarian regimes. However, here citizens can better ensure their values are being reflected in the actual procedures of government - critical given the US’s representative system.
Of course, beyond surveillance, AI may also simply provide accessibility. Much of government processes are opaque and jargon filled. AI translation could dramatically reduce the time and expertise required for citizens to understand complex policy issues, potentially enabling a more informed and engaged citizenry.
However, others cautioned that relying too heavily on AI for civic engagement could lead to a form of "outsourced citizenship," where people delegate their civic responsibilities to AI systems. If we interact with our governments primarily through AI mediators, how do we consider the AI agent’s own perspective baked in through its training? One participant expressed this concern:
"Can AI truly provide unbiased, multi-perspective information on political issues, or will it inevitably reflect certain biases?"
While this question is often presented rhetorically, we currently see organizations attempting this. And, in my own perspective, it seems like chat systems today are doing a remarkable job of hitting the right balance. I for one, am optimistic.
AI as media
However, the normal answer (and perhaps most likely) is “no”, AI will always reflect “biases”, as it must if it is to simplify the world into responses short enough to be helpful for us. Thus, rather than ask for an “unbiased” AI system, let us instead compare AI’s bias and impact to the status quo. We spoke at length about the relationship between AI and traditional media as mediator of information. If biased and oversimplified narratives are created by citizen demand rather than just a consequence of arbitrary supply, its unlikely that the shift from media to AI will have much of an impact.
To make this more concrete, imagine a future where there are many AI systems. Some strive to be “unbiased”, reflecting the news “as it is”. These systems often provide many sides of the issue and hesitate to supply simplified headlines. Current AI systems like Claude seem to take this tact. But in this future, let’s imagine another kind of AI system. These are the intentionally biased ones - they cater to a particular demographic and represent those views. These are the AI equivalent of OAN. Some participants predicted that this future would have people choose the “biased” AI systems as their preferred source. This future actually looks quite similar to today - a complicated landscape of different perspectives on the world and individual empowerment to choose your lens. In this future, AI serves as the next generation of media, just as social media and recommendation systems have changed the landscape today.
This view means that AI may have less of an impact than one may immediately think on improving our informational ecosystem. Perhaps it will instead lead to even greater filter bubbles. That said, there was still optimism. Individuals are generally curious - even the most conspiratorial people are often motivated by trying to discover “the truth”. If they are empowered with better tooling to reveal complexity in ways that are understandable, perhaps we will move to a more informed and pragmatic politics.
Ultimately, the question of whether AI serves as a tool for enhancement or a replacement for human civic engagement remains unresolved. While AI offers powerful capabilities for information processing and synthesis, it's clear that preserving human agency and fostering critical thinking skills will be crucial as we integrate these technologies into our democratic systems.
Stakeholder Alignment and Constituent Understanding
Another key theme was the challenge of aligning AI systems with the diverse and often conflicting values of multiple stakeholders in a democratic society. This challenge is particularly acute in the context of civic engagement, where the interests of individual citizens, elected officials, and various interest groups may diverge significantly.
Understanding Constituents: Do Elected Officials Need More Information?
A first step in representing multiple stakeholders is understanding them. Early in the conversation some participants assumed that AI would lead to improvements in understanding the citizenry. Through next-generation polling or automated individualized interviews, we will soon be able to understand sentiment better. Before we went too far down that road, however, the critical question was raised: Do elected officials actually need better tools to understand their citizenry? This sparked a nuanced debate among participants.
On one hand, some argued that large, well-funded campaigns already have sophisticated methods for understanding their constituents. As one participant noted:
"In a presidential campaign, they're running $100,000 in focus groups every single day, like a Senate campaign, congressional campaigns. They're spending to understand in minute detail, like, what different groups, like, think and believe and what's the driver of motivation."
This perspective suggests that, at least for major campaigns and offices, there might not be a significant information gap when it comes to understanding constituent views.
However, others pointed out that this level of resource allocation for constituent understanding is not uniform across all levels of government. As one participant observed:
"Maybe in [local races] the kind of candidates running don't have the kind of capacity to have such a sense."
The discussion then turned to how AI might help bridge this gap, particularly for smaller campaigns and local elected officials. If true, this impact of AI in politics would reflect larger trends; AI seems to raise the bottom, either by democratizing access or better augmenting those at the lower end of the performance/resource distribution. And it’s clear that in understanding individual constituents, we could do so much better. The job of the official isn’t just to poll individuals, but understand their underlying needs. Their values, their preferences - these are all important, but in a representative democracy the idealistic goal of the representative is to uncover real opportunities for government to meaningfully address citizen issues.
One participant drew a parallel to developments in the private sector:
"As a product manager, you don't want to just hear what the customers are asking for. You want to see what is their true pain. And there are many companies now that are trying to scale up user research in a more personal way by rather than interacting with a survey, you're interacting with an AI agent who is engaging."
This analogy suggests that AI could potentially provide more nuanced, interactive ways of gathering constituent feedback, going beyond traditional surveys or town halls. Such tools could be particularly valuable for local officials who lack the resources for extensive polling or focus groups.
Multi-Perspective Understanding and Decision Making
All this said, it’s possible that understanding public opinion better isn’t the main obstacle to improved democratic governance. Perhaps the main issue is just the sticky reality that in a democracy there are many incongruous beliefs that somehow need to be synthesized into a good-enough set of policies. From this perspective, polling and similar approaches flatten citizen sentiment, presenting a simplifying perspective that has been, up until now, practically necessary to inform policy. But with AI’s ability to understand and hold complexity “in mind”, perhaps these pragmatic simplifications will be less needed.
One participant highlighted the potential for AI to provide a more nuanced understanding of public opinion:
"I'm most excited about AI's capacity to hold multiple opinions at once, like, every opinion at once. In this culture of polarity, where things are one way or another way, having a system that can understand and speak to many different rationales is really exciting."
This multi-perspective approach could potentially help bridge partisan divides and facilitate more constructive political discourse. However, it also raises questions about how to weight and balance these diverse perspectives in decision-making processes.
Which brought us to power dynamics. The power dynamics between different stakeholders emerged as a critical factor in determining how AI systems might be developed and deployed in civic contexts. Participants highlighted that commercial interests both impact AI development and governance, and that idealistic notions that AI will be in any more service to the citizenry than government was naive.
Conclusion: Navigating the AI-Enabled Civic Landscape
As our discussion revealed, the integration of AI into civic processes presents both immense opportunities and significant challenges, as it is in almost all of our discussions. The key issues we explored - the role of AI as a tool or replacement for human engagement, and the challenge of multi-stakeholder alignment - are likely to be central to discussions about AI and democracy for years to come.
The path forward will require careful navigation of these complex issues, balancing the potential of AI to enhance civic engagement with the need to preserve human agency and democratic values. As one participant optimistically concluded:
"I really think we can bring down the barrier to entry for meaningful and agentic information. And that makes me extremely optimistic."
As we continue to develop and deploy AI systems in civic contexts, it will be crucial to maintain ongoing dialogue between technologists, policymakers, and citizens. Only through such collaborative efforts can we hope to harness the power of AI to create more responsive, inclusive, and effective democratic systems while safeguarding the core values that underpin our societies.
Notes from the conversation
There's a tension between the need for civic education and the practical challenges of engaging citizens in complex policy issues.
AI could potentially lower barriers to civic engagement by making complex information more accessible and digestible.
The current political system often prioritizes money and special interests over broader civic participation.
There's debate over whether a "civic elite" is beneficial or if civic engagement should be more broadly democratized.
Evidence-based policymaking is desirable but challenging to implement due to complexities in measuring policy impacts.
Direct democracy enabled by technology is seen as promising by some, but others worry about the public's capacity to engage deeply on complex issues.
AI could potentially help aggregate public opinions in more nuanced ways than current polling and voting systems.
There are concerns about AI amplifying misinformation and polarization in political discourse.
Some see potential for AI to provide more balanced, multi-perspective information on political issues.
Time constraints and economic pressures are major barriers to civic engagement for many people.
There's debate over whether AI could serve as an impartial "philosopher king" or if human decision-making should remain primary in governance.
The "Overton window" of acceptable political discourse could potentially be shifted by AI-enabled information sharing and opinion aggregation.
Some argue current representative democracy doesn't adequately reflect constituent preferences, especially those of less wealthy citizens.
AI assistants could potentially help citizens better understand complex legislation and policy proposals.
There are varying views on whether broader participation or a more engaged "civic elite" would lead to better governance outcomes.
Some see potential for AI to make government operations and decision-making more transparent to citizens.
Cultural shifts may be needed alongside technological solutions to truly increase civic engagement.
There's interest in using AI to provide personalized, relevant civic information to individuals.
Some argue current political polarization stems partly from people lacking diverse perspectives, which AI could potentially help address.
There are concerns about stratification of information sources, even with AI, leading to further polarization.
Questions
How can we balance the need for informed civic participation with the reality that most citizens have limited time and resources to deeply engage in policy issues?
Can AI truly provide unbiased, multi-perspective information on political issues, or will it inevitably reflect certain biases?
How might widespread use of AI in civic engagement shift the balance of power between citizens, elected officials, and special interest groups?
Is direct democracy enabled by AI desirable, or are there benefits to representative systems that shouldn't be lost?
How can we ensure AI-enabled civic tools don't exacerbate existing inequalities in political influence?
What are the potential risks of relying too heavily on AI for interpreting and simplifying complex policy information?
How might AI-enabled civic engagement tools change the nature of political campaigns and lobbying?
Can AI help bridge partisan divides, or is there a risk it could further entrench people in their existing views?
How do we balance making civic information more accessible while ensuring citizens still develop critical thinking skills about political issues?
What are the implications of potentially having different AI "agents" representing different political viewpoints?
How might the role of elected officials change if constituents have AI tools giving them deeper policy insights?
Could AI-enabled civic tools lead to more rapid policy changes, and what are the potential benefits and drawbacks of this?
How do we ensure transparency in how AI civic tools are developed and deployed to maintain public trust?
What are the ethical considerations around using AI to personalize political information for individuals?
How might AI change the way we measure public opinion and the impacts of policies?
Could AI-enabled civic engagement lead to a more educated and engaged citizenry, or risk creating a false sense of understanding?
How do we balance using AI to simplify complex issues while still conveying important nuances?
What are the potential national security implications of using AI extensively in civic and political processes?
How might AI impact local vs. national level civic engagement differently?
Could over-reliance on AI in civic processes ultimately undermine human agency in democratic decision-making?