TLDR: This research paper, “Levers of Power in the Field of AI: An Ethnography of Personal Influence in Institutionalization,” explores how decision-makers in academia, government, business, and civil society exercise and experience “levers of power” to influence AI implementation. Through twelve fictional personas based on real questionnaire responses, the study illustrates how personal agency, organizational structures, and institutional infrastructures intersect in AI governance. It discusses the relative personal power of decision-makers, strategies for fostering institutional stability during change, and methods for influencing institutional change, concluding with five testable hypotheses on the dynamics of power in the AI field.
The rapid advancement of Artificial Intelligence (AI) presents a complex landscape for decision-makers across various sectors. A recent study, titled “Levers of Power in the Field of AI: An Ethnography of Personal Influence in Institutionalization,” delves into how individuals in academia, government, business, and civil society navigate the intricate questions of power when implementing AI. This research introduces the concept of “levers of power” as social mechanisms through which individuals can shape institutional responses to technological change.
The study, conducted by Tammy Mackenzie, Sukriti Punj, Natalie Perez, Sreyoshi Bhaduri, and Branislav Radelji´c, employed an ethnographic research design. It utilized personalized questionnaires to gather insights into decision-makers’ institutional roles, drawing on a governance framework rooted in Neo Institutionalism and extended with the dimension of “Idea Mobility.” The findings are presented through twelve fictional personas, each representing a high-level decision-maker from North America and Europe. These personas, while anonymized, are based on real responses and circumstances, illustrating the interplay of personal agency, organizational logics, and institutional infrastructures in AI governance.
Understanding the Levers of Power
The research identifies several key “levers of power” that individuals can utilize within an “institutional field” – a system encompassing all practices related to a given topic, such as AI in society. These levers range from formal and systemized to informal and diffuse. They include: logics (e.g., market, social justice), the elaboration of institutional infrastructure, governance structures, collective interest organizations, regulators, informal governance bodies, field-configuring events, status differentiators, organizational models, categories/labels, norms, relational channels, and idea mobility.
Insights from Decision-Makers
The personas reveal diverse perspectives and experiences. For instance, Alex, a policymaker in academia, focuses on generating policy options with input from various specialists and identifies international policy conferences as significant field-configuring events. Cameron, a senator, highlights the strong influence of established practices in government and the necessity of engaging with collective interest organizations, though with a critical eye. Dana, an architect in Responsible AI in business, aims to solve social problems through literacy programs and defining technical requirements, combining humanitarian aid with formal AI ethics education.
Emerson, a senior rabbi and community leader, exemplifies adaptability, recounting how the pandemic forced a reimagining of centuries of tradition, moving religious services online. This illustrates how external upheavals can drive institutional change and foster new dependencies. Finley, an academic researcher, expresses concern about students misusing AI, such as writing PhD theses with ChatGPT, emphasizing the need to empower AI use without breaking rules. Kinsey, an executive director of a new think tank, underscores the importance of informal norms and personal conviction in exercising social power. Lindsay, a scientific director, voices a profound concern about humanity collectively racing forward with AI without knowing how to control a potentially super-intelligent system.
Personal Power and Institutional Change
The study found that many decision-makers feel empowered to effect change within their organizations, even in tradition-bound institutions. This sense of empowerment often precedes the successful exercise of power. Interestingly, individuals in more complex institutions, like national governments, reported similar levels of personal empowerment despite perceived less direct influence. The paper also notes that the logic guiding a decision-maker’s actions may not always align with their sector; for example, a business executive might act from a logic of social good, while an academic considers cost-effectiveness. This suggests that personal relationships and tailoring rationales to individual preferences are crucial for influencing change.
A notable contradiction emerged regarding accessibility: while most participants claimed stakeholders could reach them, the reality of societal disenfranchisement and structural constraints often makes public engagement difficult. The persona of Jaime, a public servant, illustrates this tension, emphasizing the need for anonymity for ethical and security reasons while also affirming accessibility. This highlights a “decoupling” between stated rules and actual practice, where institutional insiders must find mechanisms to resolve such conflicts.
Also Read:
- Navigating the AI Divide: A New Framework to Measure Policy-Practice Gaps in Academia
- Navigating AI Risks: Understanding Diverse Perspectives with LLMs
Hypotheses for Future Research
Based on their analysis, the authors propose five testable hypotheses about the dynamics of these levers of power in the AI field:
- Formal methods are outpaced, with increasing leverage for informal methods.
- Institutional dynamics are dominated by business marketing and community disengagement.
- Formal and informal power mechanisms are being strategically addressed by big tech.
- Informal influence remains elitist and siloed.
- Collective action is growing rapidly through large-scale initiatives.
In conclusion, this research offers valuable insights into how individuals can personally engage with AI governance, emphasizing the importance of understanding power dynamics and the pressures on institutional change. As one scientific director eloquently put it, “We do not know yet how we would control AI that could be smarter than us, yet we collectively race forward, taking an existential risk for humanity’s future.” This study serves as a bridge for policymakers and civil society advocates to navigate the complexities of AI and work towards a future where technology serves the greater good. For more detailed information, you can read the full research paper here.
