AI and Politics – as in practical campaigning, not as in producing Deep Fakes

Politico published this article about AI with the note, “This article is part of a series, Bots and ballots: How artificial intelligence is reshaping elections worldwide, presented by Luminate.” (1)

Luminate is a consortium that supports various media projects. “It was established by philanthropists Pierre and Pam Omidyar and has worked for over a decade on issues related to governance and citizen engagement.” (2)

The article is about a practical application of the latest development AI in politics. I mention its connection to the Luminate consortium because with the hype and the fantasy and fears that are so much a part of the current publicity around AI, it’s important to keep a focus on the fact that we are dealing with a technology (artificial intelligence) that has been with us for decades and continues to develop.

This is a short report from Yahoo Finance from last December on this topic (3):

Part of the recent attention to AI has focused on its capabilities for creating Deep Fakes, a real concern in politics generally and specifically in the field of government information operations, aka, propaganda operations.

But it’s also very misleading to think of AI as the forerunner of SkyNet, the evil autonomous computer system of the Terminator universe. It has much more practical and less terrifying usages that producing sinister Deep Fakes for Russian disinformation operations.

In the Politico report, Mark Scott interviews Hannah O’Rourke, who is an advocate of the use of AI to empower voters:

Throughout monthly hackathons, O’Rourke and other progressive computer scientists at CampaignLab, a nonprofit that she co-founded, whipped up an AI-powered chatbot. Designed with different personalities and varying emotions, it helps volunteers learn how to best interact with potentially skeptical voters on the campaign trail.

Another crowdsourced project relied on AI tools to track political discussions on TikTok. A third used off-the-shelf AI technology to help confused voters decipher the labyrinthine website of the country's election commission.

"There are some interesting, creative AI solutions that can help humans do better at things like campaigning," O’Rourke said over coffee last month on a sweltering spring day in South London — with kids laughing at a nearby playground and commuters grabbing a drink on their way home from work. [my emphasis] 

This is a practical application that is neutral in the sense that it can be adapted for very different kinds of causes. It can be used to help protesters organize against war or in favor of better environmental protections. It can also be used to refine arguments in favor of disastrous military action or to campaign for looser anti-corruption laws.

Alexander Brentler recently summarized the state of the AI “large language models,” of which chatbots like ChatGPT and Google Gemini are examples:

There's a reason for the enormous concentration of capital in AI research: GPT-4 seems to effortlessly master formalized written communication. The texts of which you have seen enough examples – a restaurant review, a sonnet, a love letter – are not only perfectly reproduced by the model, but also easily adapt in style and content according to our wishes. This also applies to "structured" texts, such as program code or moderately difficult mathematical tests. In addition, the latest generation of large language models has a distinctly playful character. In words and images, they capture parody and satire almost perfectly. The latest language models continue to make mistakes and often fail at surprisingly simple tasks, but there's no denying that great progress has been made. (4)

Training political canvassers to respond to questions they should expect to encounter is a plausible usage of a chatbot.

I’ve pretty much always been a big fan of Herbert Marcuse’s One-Dimensional Man (1964), so I’m going to stop short of saying that the technology is perfectly ideological neutral, even though it can be used to construct arguments for and against the same position. But you could ask a chatbot to take make arguments based on Marcuse’s book for and against the concept that large language models are ideologically neutral. I haven’t tried that myself yet. But based on my admittedly limited experience, that would be less a trust-but-verify exercise than an oh-boy-you-really-need- to-verify-this´-before-you-use-it.

But I’m reproducing below an experiment I did do, asking ChatGPT about one of the most crackpot but nevertheless still-popular claims of “supply-side” economics, to whose dubious blessings St. Reagan as President introduced the broader American public – much to the disadvantage of the large majority.

Since I’ve been engaging with the claims of supply-side economics longer than I care to think, I can comfortably say that the AI bot’s response was decent. If I were preparing for a live discussion on the subject, it’s not a bad start. But I would then also take the step of asking the chatbot to refute the arguments it provided, to give me an idea of what kind of challenges my arguments might face. I wish I had had something like this available in my days as a high-school and college debate-contest days!

It's also notable that the response doesn’t cite sources (and I didn’t include citing sources in the query). But if I were going to use this to build my own argument, I would flesh it out with backup sources. But this ChatGPT response is certainly a decent starting point for that. The prose is a bit flat but the content itself is actually clear.

My ChatGPT query on supply-side economics

I asked ChatGPT: “In five hundred words or less, refute the argument that in the United States cutting tax rates for the highest tax brackets always produces more economic growth.”

It produced the following response, which I’m quoting in full here:

The argument that cutting tax rates for the highest income brackets always produces more economic growth in the United States is flawed for several reasons. While tax cuts can sometimes stimulate the economy, the relationship between tax rates and economic growth is complex and influenced by various factors.

Firstly, historical evidence does not consistently support the claim that tax cuts for the wealthy lead to significant economic growth. For example, during the post-World War II period, the United States experienced robust economic growth despite high marginal tax rates. In the 1950s and 1960s, the top marginal tax rate was above 90%, yet the economy expanded rapidly, wages increased, and the middle class grew.

Secondly, tax cuts for the wealthy can lead to increased income inequality, which can have adverse effects on economic growth. High levels of inequality can reduce aggregate demand because the wealthy tend to save more of their income compared to lower-income individuals, who are more likely to spend additional income. This reduced consumption can slow economic growth. Studies have shown that a more equitable income distribution can support stronger and more sustainable economic growth by fostering greater consumer spending and investment in human capital.

Additionally, tax cuts for the highest earners can lead to budget deficits and reduced public investment. When tax revenues decline, the government may need to cut spending on essential services and infrastructure, such as education, healthcare, and transportation. These public investments are crucial for long-term economic growth, as they enhance the productivity and competitiveness of the workforce. Reducing funding for these areas can hinder economic progress and reduce opportunities for future generations.

Moreover, the assumption that the wealthy will invest the additional income from tax cuts in ways that significantly boost the economy is not always valid. While some of the extra income might be invested productively, much of it may go into savings, luxury consumption, or speculative investments that do not contribute to broader economic growth. The effectiveness of tax cuts in stimulating investment depends heavily on the economic context and the availability of profitable investment opportunities.

Finally, targeted tax cuts or increased public spending directed towards lower and middle-income individuals can be more effective in stimulating economic growth. These groups are more likely to spend additional income on goods and services, which can boost demand and drive economic activity. Programs like the Earned Income Tax Credit (EITC) or increased infrastructure spending can have a more direct and substantial impact on economic growth by supporting those who are more likely to spend and by creating jobs.

In conclusion, while tax cuts for the highest income brackets may provide some economic stimulus under certain conditions, they do not always lead to more economic growth and can have significant downsides. Historical evidence, the potential for increased inequality, the risk of reduced public investment, and the uncertain impact on productive investment all suggest that a more nuanced approach to fiscal policy is necessary. A balanced mix of tax policies and public investment that supports broad-based economic growth is more likely to yield sustainable economic benefits for the country as a whole.

Notes:

(1) Scott, Mark (2024): AI, Inc. flexes its election bona fides — and hunts for customers Politico 05/21/2024, <https://www.politico.eu/article/artificial-intelligence-ai-elections-companies-campaigners-voters-bots-ballots/> (Accessed: 2024-25-05).

(2) Luminate, n/d. The Omidyar Group. (Accessed: 2024-25-05). <https://www.omidyargroup.com/pov/organizations/luminate/> (Accessed: 2024-25-05).

(3) Generative AI: Impact on 2024 election. Yahoo Finance YouTube channel. <https://youtu.be/YXNpi8X_LzI?si=_KbSH7slFnSG8mSN> (Accessed: 2024-25-05).

(4) Brentler, Alexander (2024): La Inteligencia Artificial resuelve problemas inexistentes. Jacobin latino 04.04.2024. <https://jacobinlat.com/2024/04/04/la-inteligencia-artificial-resuelve-problemas-inexistentes/> (Accessed: 2024-25-05). I used the Microsoft Word translation function – which is also an AI function - to assist in this translation from Spanish. The Word/AI translation was very good but not entirely optimal in my view, so I made some adjustments. If someone thinks the translation is suboptimal, you have to blame me and not the Word translation function. AI translations are a trust-but-verify kind of thing.