News

With my wife, Jitka Špeciánová, we have published a chapter in the new book Proč se nebát AI (Why Not Worry About AI?). In the text, we explore the complex relationship between AI and labor markets. 

While many fear that AI will lead to widespread job losses, we maintain a cautiously optimistic view based on historical precedent - technological revolutions have typically created new opportunities alongside disrupting existing jobs. However, we acknowledge that this "fourth industrial revolution" may be fundamentally different from previous ones, as AI targets intellectual and high-skilled work rather than just physical labor. 

Though challenges lie ahead, particularly in potential technological unemployment, we believe human ingenuity and institutional adaptations (such as universal basic income) can help us navigate these changes successfully. Our key message is that while the challenges posed by AI shouldn't be taken lightly, human resourcefulness and adaptability have consistently helped us overcome technological disruptions in the past.

At TEDA’24: The Magic Machine, Through the Prism of Art and Science conference, held at the University of Cambridge, Lucy Císař Brown and I presented our research on "AI Oracles: Technological Re-enchantment of the World." Our talk explored the evolving role of AI in society, proposing a shift in how we perceive and interact with this technology.

Key Points:

1. AI as Oracle: We argued that AI systems are transitioning from tools to "digital Oracles," fundamentally changing our relationship with technology.

2. Technological Re-enchantment: We proposed that AI's rise marks a reversal of Max Weber's concept of societal disenchantment, instead leading to a form of re-enchantment centered on AI.

3. AI's Mystique: The complexity and opacity of AI systems contribute to their Oracle-like status. Even for experts, the processes between input and output often remain mysterious.

4. Risks and Challenges: While acknowledging AI's potential benefits, we pointed out risks such as increased isolation and the erosion of shared rational discourse.

5. Quest for Meaning: As AI takes on more and more tasks, we emphasized the challenge of maintaining human agency and finding meaning in an AI-mediated world.

At AmKon 2024, a student simulation of the U.S. Congress held in Pilsen, I delivered a lecture on "Large Language Models in Democratic Society." My talk provided an overview of the current state and potential future of generative AI, particularly large language models (LLMs), and their impact on democratic processes.

I discussed the rapid development of generative AI, highlighting its evolution from simple text-based models to more advanced systems capable of processing and generating various media types. I emphasized the potential of these technologies to both challenge and enhance democratic institutions.

A key focus of my presentation was the concept of "digital advisors" - the use of LLMs as assistants in representative bodies. I explored the advantages of such systems, including immediate access to information and the ability to process large amounts of data, while also addressing potential risks such as misinformation and value imposition.

I concluded the talk by considering how democratic mechanisms might adapt to incorporate these technologies effectively, stressing the importance of maintaining critical debate, diverse perspectives, and innovative approaches to information verification in the digital age.

I spoke at the Conference on Educational Activities of Universities, held on September 5-6, 2024, at the Faculty of Architecture of the Czech Technical University in Prague. The event, which focused on learning outcomes in higher education, was organized under the auspices of the Czech Ministry for Education, Youth and Sports.

My keynote presentation, titled "Generative AI and the Future of Tertiary Education," addressed the challenges and opportunities that generative AI presents for higher education. In my talk, I discussed potential changes in curriculum and teaching methods, as well as the need for institutional support to adapt to this rapidly evolving technology.

I emphasized the importance of universities taking an active stance in addressing these developments to maintain educational effectiveness and societal relevance. My aim was to highlight how crucial it is for educational institutions to engage proactively with these new technologies to ensure they continue to provide valuable and relevant education in our changing world.

My new preprint "Digital Homunculi: Reimagining Democracy Research with Generative Agents" is now live on arXiv.

Abstract: The pace of technological change continues to outstrip the evolution of democratic institutions, creating an urgent need for innovative approaches to democratic reform. However, the experimentation bottleneck - characterized by slow speed, high costs, limited scalability, and ethical risks - has long hindered progress in democracy research. This paper proposes a novel solution: employing generative artificial intelligence (GenAI) to create synthetic data through the simulation of digital homunculi, GenAI-powered entities designed to mimic human behavior in social contexts. By enabling rapid, low-risk experimentation with alternative institutional designs, this approach could significantly accelerate democratic innovation. I examine the potential of GenAI-assisted research to mitigate current limitations in democratic experimentation, including the ability to simulate large-scale societal interactions and test complex institutional mechanisms. While acknowledging potential risks such as algorithmic bias, reproducibility challenges, and AI alignment issues, I argue that the benefits of synthetic data are likely to outweigh their drawbacks if implemented with proper caution. To address existing challenges, I propose a range of technical, methodological, and institutional adaptations. The paper concludes with a call for interdisciplinary collaboration in the development and implementation of GenAI-assisted methods in democracy research, highlighting their potential to bridge the gap between democratic theory and practice in an era of rapid technological change.

After 12 amazing years at the Faculty of Humanities, I am joining the faculty at the Institute of Political Sciences of Charles University's renowned Faculty of Social Sciences. 

I will also serve as the Coordinator for Student Affairs at the quickly growing Politics, Philosophy and Economics program (PPE Prague).

Exciting times!

My paper "Machine Advisors: Integrating Large Language Models Into Democratic Assemblies" was just published in Social Epistemology: A Journal of Knowledge, Culture and Policy. You may access the gated version here or just download the unformatted manuscript by clicking on the pdf to the left.

Abstract: Could the employment of large language models (LLMs) in place of human advisors improve the problem-solving ability of democratic assemblies? LLMs represent the most significant recent incarnation of artificial intelligence and could change the future of democratic governance. This paper assesses their potential to serve as expert advisors to democratic representatives. While LLMs promise enhanced expertise availability and accessibility, they also present specific challenges. These include hallucinations, misalignment and value imposition. After weighing LLMs’ benefits and drawbacks against human advisors, I argue that time-tested democratic procedures, such as deliberation and aggregation by voting, provide safeguards that are effective against human and machine advisor shortcomings alike. Additional protective measures may include custom training for advisor LLMs or boosting representatives’ competencies in query formulation. Implementation of adversarial proceedings in which LLM advisors would debate each other and provide dissenting opinions is likely to yield further epistemic benefits. Overall, promising interventions that would mitigate the LLM risks appear feasible. Machine advisors could thus empower human decision-makers to make more autonomous, higher-quality decisions. On this basis, I defend the hypothesis that LLMs’ careful integration into policymaking could augment democracy’s ability to address today’s complex social problems.

I presented my research on "Beyond the Limits: Generative AI and Synthetic Data in Democracy Research" at the ECPR General Conference 2024 in Dublin. In my talk, I explored the potential of generative AI and synthetic data to accelerate evidence-based democratic innovation. I discussed the current challenges in democracy research, introduced the concept of digital homunculi, and examined both the promise and pitfalls of using AI-generated synthetic data in social science research. I also suggested technical, methodological, and institutional adaptations to integrate these new approaches into democracy studies. #ecprgc24

At the 7th International Conference Economic Philosophy in Reims, I presented a talk on "Generative AI in the Marketplace of Ideas"  revolving around the hypothesis that the rise of GenAI marks a pivotal moment for the marketplace of ideas with highly significant long-term consequences for democracy.

I am also very grateful to the organizers, Cyril Hedoin and Jean-Sébastien Gharbi, for dedicating a session to my book Behavioral Political Economy and Democratic Theory.  Many thanks to Malte Dold (Pomona College), Otto Lehto (NYU), and Andres Moles (CEU) for their excellent contributions!

Martin Buchtík and Klára Zajíčková from STEM Institute for Empirical Research invited me to talk about LLM chatbots, their use cases in social scientific research, and our AI Institutional Transformation Research Group. You may listen to the podcast here (in Czech). 

The Institute for Information Studies and Librarianship (Charles University) has invited me to deliver a lecture on democracy in the digital age.  My lecture explores the current challenges faced by liberal democracy, which has previously enabled unprecedented growth in human prosperity. The main issues addressed are the rise of political polarization and authoritarianism. The lecture examines these developments through the lens of behavioral political economy, questioning the role played by information technology in this context. With the rapid emergence of artificial intelligence, understanding this question has become particularly urgent. The lecture takes place against the backdrop of these technologies' increasing influence on democratic processes and discourse, seeking ways to harness the power of technology to facilitate democratic innovations.

Alena Ivanova has interviewed me about my research and views on democracy in the digital age and the expected impacts of artificial intelligence. You can find the interview here (in Czech). 

Here is how Claude 3 summarizes the content:

Petr Špecián, founder of the AI Institutional Transformation Research Group, says his career trajectory changed forever after attending philosophy lectures by chance while studying international trade. His research now focuses on how institutions adapt to new technologies like large language models and their practical applications, as well as the relationship between democracy and expertise in the digital age.

Špecián notes that technological development is moving extremely rapidly compared to institutions, which adapt more slowly, creating tension. He says liberal democracy developed stably when mass media was the broadly shared information source, but the proliferation of decentralized information sources today makes it harder for people to agree on a shared reality, putting pressure on democracies.

On whether AI benefits or threatens democracy, Špecián sees a range of possible scenarios, from finding new norms leading to better democracy, to AI enabling authoritarianism or chronic stagnation. He emphasizes that AI is not value-neutral, as the corporations developing it imbue it with progressive values not everyone shares.

Špecián is interested in whether elected representatives could use AI chatbots to replace expert advisors and better ensure democratic values shape collective decisions. While AI still "hallucinates" false information, he sees good prospects for improvement. He believes people will adapt to AI-generated disinformation, though it's unclear if solutions will be centralized or decentralized.

Overall, Špecián advocates for personal experience using advanced AI systems to understand their powerful capabilities. He appreciates experimentation with alternatives to standard approaches, like replacing bachelor's theses, in the face of revolutionary technological change.

My new mini-paper "Give the machine a chance, human experts ain't that great…", co-authored with Lucy Císař Brown, is now published in AI & Society. Fulltext available here: https://t.co/MPVASqsuOC.

We argue that, despite their flaws, large language models (LLMs) deserve a fair chance to prove their mettle against human experts, who are often plagued with biases, conflicts of interest, and other frailties. For epistemically unprivileged laypeople struggling to access expert knowledge, the accessibility advantages of LLMs could prove crucial. While concerns about LLMs' inconsistencies and arguments for human superiority are often justified (for now), they may distract from the urgent need to prepare for the likely scenario of LLMs' continued ascent. Experimentation with both the capabilities and institutional architecture of LLMs is necessary. As LLMs are here to stay and they keep improving, it is high time we started thinking about how to navigate the impending wave of their proliferation.

My preprint "Machine Advisors: Integrating Large Language Models into Democratic Assemblies is now available via SSRN.

Abstract: Large language models (LLMs) represent the currently most relevant incarnation of artificial intelligence with respect to the future fate of democratic governance. Considering their potential, this paper seeks to answer a pressing question: Could LLMs outperform humans as expert advisors to democratic assemblies? While bearing promise of enhanced expertise availability and accessibility, they also present challenges of hallucinations, misalignment, or value imposition. Weighing LLMs’ benefits and drawbacks compared to their human counterparts, I argue for their careful integration to augment democracy’s ability to address complex policy issues. The paper posits that time-tested democratic procedures like deliberation and aggregation by voting provide safeguards effective against both human and machine advisor imperfections. Additional protective measures include custom LLM training for the advisory role, boosting representatives’ competencies in query formulation, or implementation of adversarial proceedings in which LLM advisors could debate each other and provide dissenting opinions. These could further mitigate the risks that LLMs present in advisory role and empower human decision-makers toward increased autonomy and quality of their collective choices. My conceptual exploration offers a roadmap for the co-evolution of AI and democratic institution, setting the stage for an empirical research agenda to finetune the implementation specifics. 

My lecture "Social Technologies of the Future" and a podcast on the social consequences of AI (with Jaku Krejčí as an inreviewer) are now available online. They have been recorded on a summer school organized bz University of Hradec Králové.

Lecture annotation: Under the term technology, we are accustomed to imagining manufacturing processes that produce material goods: for example, cars, computers, or industrial robots. However, in a broader sense, this term also includes social technologies, such as capitalism and liberal democracy - that is, ways in which human society is organized.

With the development of material technologies, which today leads to an increasingly extensive digitalization of the economy and political discourse, existing social technologies may encounter the limits of their usefulness and adaptability. Capitalism and liberal democracy, as we know them, could become outdated due to digitalization, just as previous social forms became outdated due to previous industrial revolutions.

What could the social technologies of the future look like, which will better cope with technological unemployment, decentralized information flows, or the progressing dematerialization of economic goods? I will review several existing proposals, such as the introduction of unconditional basic income or democracy based on sortition instead of elections, and discuss their strengths and weaknesses. Based on this discussion, the possible social technologies of the future will be shown in more precise contours.

Our research project led by Miroslav Vacura with team members Petr Špecián and Eugenia Stamboliev has obtained funfing for 2024-2026 from the Czech Science Agency (GACR). The project engages in a philosophical inquiry into the role and impact of Large Language Models (LLMs) within democratic societies. In an era marked by rapid technological advancements, this research aims to critically assess the epistemic consequences of LLMs on democratic discourse, the epistemic authority of expertise, and the legitimacy of democratic institutions. Central to our study is the exploration of the potential effects of LLMs in democratic settings and strategies to ensure their alignment with democratic values. The project will delve into the ethical responsibilities incumbent upon governmental bodies and citizens in shaping and utilizing LLMs, reflecting on the intersection of technology, democracy, and ethics.

I have presented a talk on "Reimagining Institutions through Generative AI" at the conference Meaning of Democracy organized by the Polish Academy of Sciences - Scientific Centre in Vienna, Austria.

Abstract: The talk focuses on the potential of using large language models (LLMs) as a means to strengthen democracy in the face of modern challenges. Petr argues that the integration of AI, specifically LLMs, into democratic processes can enhance collective intelligence and decision-making. LLMs, with their ability to process vast amounts of information and provide coherent responses, could surpass the limitations of human experts who are often constrained by scarcity of resources, conflicts of interest, and narrow focus. The proposed approach includes implementing institutional safeguards and starting with experimental use of LLM advisors in controlled environments to gather empirical evidence. The main message is optimistic, suggesting that the strategic use of AI can empower democracies to effectively manage contemporary issues and complexities.

I have presented a talk on "Reimagining Institutions through Generative AI" at Computing the Human Conference in Brno.

Abstract: As AI systems grow more advanced, new possibilities emerge to utilize their generative capabilities to reimagine our institutions. It is becoming possible to integrate technology with the social sciences to develop AI-powered capabilities for institutional innovation. My talk will focus on the intriguing possibility of constructing a collaborative digital sandbox where researchers could rapidly prototype and simulate the impacts of new institutional architectures. This virtual institutional laboratory—or playground—would provide a space to experiment with novel designs optimized to answer the 21st century’s challenges. Interdisciplinary teams of social scientists, computer scientists, and government innovators would co-create this platform and populate it with imaginative institutional alternatives to be tested and refined “in silico” before their real-world deployment. Thus, when carefully directed, AI’s generative power could help design new forms of social organization and governance fit for the digital age. 

My talk on "Democracy in Digital Age" from the 2022 Být nebo bit conference in now available on YouTube.

The talk covers the contemporary troubles of the world's democracies, the role political irrationality plays in them, and the possible cures. 

I presented a talk "Democratizing Expertise. The Limits and Possibilities of Artificial Intelligence in Enhancing Democratic Decision-Making" at the ECPR General Conference 2023.

In the talk, I delved into the potential of integrating large language models (LLMs) like GPT-4 into democratic decision-making. I acknowledged their limitations, such as opacity and the potential for value imposition but argued that these aren't necessarily more problematic than the shortcomings inherent in relying on human experts. I concluded that, if managed responsibly, LLMs could become a vital asset in democratizing expertise and thereby enhancing democratic processes.

I presented a talk "Language Models for Democracy: Limits and Possibilities" on Technology &  Sustainable Development 2023 conference at Østfold University College, Norway.

I have received the great privilege and responsibility to have been appointed a Vice-Dean for Digitalization at Faculty of Humanities, Charles University. 

Together with Filip Tvrdý, Petra Chudárková, and Miroslav Vacura we organized an interdisciplinary conference on the challenges democracies face in the digital age. We held 15 talks by speakers from 7 countries and teckled the issue from several angles. My talk Large Language Modes for Democracy has been a part of the program.

Special thanks to Petr Večeř for his outstanding assistance with the conference organization.

Together with my colleagues from the Faculty of Humanities, Charles University, I founded an AI Institutional Transformation Research Group. We are an interdisciplinary team of scholars committed to the study of institutional changes driven by the rapid progress and spread of the technologies of artificial intelligence in the form of generative models, such as GPT-4 or Stable Diffusion.