Systemic Approach and Leverage Points in AI Policies. Part I

The views expressed are those of the author and do not necessarily reflect the views of ASPA as an organization.

By Mauricio Covarrubias
December 13, 2024

In a world increasingly driven by artificial intelligence (AI), public policies face the dual challenge of maximizing the benefits of this technology while mitigating its risks. AI systems operate in highly complex technological, social and economic environments, giving rise to challenges such as algorithmic biases, technological inequalities, and unintended side effects.  Addressing these issues requires a systemic approach and the strategic use of leverage points—specific areas within a system where relatively small interventions can produce significant and lasting impacts (Meadows, 1999).

The Impact of Artificial Intelligence on Complex Systems

Artificial intelligence is profoundly transforming sectors such as healthcare, education, the economy and justice, thanks to its capacity to process large datasets, identify hidden patterns and automate decision-making. These capabilities make it a powerful tool for addressing complex problems that have traditionally been difficult to manage. However, the integration of AI into these sectors is not without risks, particularly when applied to complex systems characterized by interdependence and unpredictability.

A notable example is the use of predictive algorithms in the judicial system to estimate recidivism rates and guide decisions on parole. While these tools aim to enhance efficiency and objectivity, research by Angwin et al. demonstrates that such algorithms often perpetuate preexisting biases. Marginalized communities frequently bear the brunt of these biases, as historical data used to train algorithms reflects systemic inequalities. This not only results in unjust outcomes but also erodes public trust in the judicial system.

The complexity of systems where AI is deployed lies in their interdependence. A change in one component—such as modifying an algorithm or the data that feeds it—can produce ripple effects throughout the system. Additionally, the cumulative and often invisible nature of AI’s impacts exacerbates this complexity. Automated decisions not only create immediate consequences but also shape the system’s long-term trajectory, amplifying patterns or inequities over time. This makes it challenging to anticipate and address the full scope of AI’s consequences, especially when they become evident only after significant harm has occurred.

Leverage Points: Prioritizing Strategic Interventions

Donella Meadows introduced the concept of leverage points as specific areas within a system where targeted interventions can lead to significant and systemic change. This framework is particularly relevant in the context of AI, where systems are often vast, interconnected and resistant to superficial fixes.  Leverage points vary in depth and effectiveness, ranging from shallow adjustments, such as tweaking parameters, to profound transformations, such as altering the paradigms that guide a system’s behavior.

Identifying and prioritizing these points is crucial for crafting AI policies that address both the symptoms and root causes of systemic challenges.

Shallow Leverage Points: Parameter Adjustments and Performance Metrics

At the surface level, adjusting technical parameters, such as changing performance metrics, is one of the easiest interventions. For example, replacing a metric like “accuracy” with one that prioritizes fairness can reduce algorithmic bias in specific contexts. While these changes are relatively simple and can yield quick results, they often fail to address deeper structural and systemic issues. For instance, shifting performance metrics does not eliminate the inequalities embedded in training data or address broader socio-economic disparities.

Policymakers must view shallow leverage points as a starting point rather than a solution. These adjustments can be effective when paired with broader interventions that address the structural and cultural dimensions of AI systems.

Intermediate Leverage Points: Structural and Regulatory Changes

Structural interventions involve modifying the rules and mechanisms governing a system. Examples include introducing regulations, mandating transparency or requiring oversight of AI systems. For instance, the European Union’s Artificial Intelligence Act sets standards for high-risk AI applications, requiring audits, ethical compliance and penalties for violations. These rules not only improve current practices but also establish a precedent for future AI technologies.

Structural leverage points are critical for addressing systemic risks that arise from unregulated AI deployment. However, their effectiveness depends on adaptability and enforcement. Policymakers must ensure that these interventions are flexible enough to evolve alongside rapid technological advancements while remaining robust in safeguarding public interests.

Deep Leverage Points: Transforming Paradigms

At the deepest level, leverage points involve changing the paradigms that shape a system’s purpose and behavior. In the case of AI, the prevailing paradigm often prioritizes efficiency, economic growth and technological progress. While these goals have driven innovation, they have also exacerbated social inequities and ethical dilemmas.

Transforming this paradigm to emphasize equity, sustainability and human rights could fundamentally alter the trajectory of AI development. Initiatives like the Montreal Declaration for Responsible AI advocate for centering people in the development of AI technologies. By embedding values such as fairness and accountability into the design of AI systems, policymakers can ensure that these technologies serve societal needs rather than purely economic interests.

Paradigm shifts require long-term commitment and active engagement with diverse stakeholders, including governments, civil society and affected communities. They also demand a reevaluation of societal values to align technological innovation with the broader public good.

In the next part, we will delve into the practical application of these concepts, exploring how leverage points can transform complex systems.

---

Author: Mauricio Covarrubias is Professor at the National Institute of Public Administration in Mexico.  He is co-founder of the International Academy of Political-Administrative Sciences (IAPAS).  He is the founder and Editor of the International Journal of Studies on Educational Systems (RIESED). Member of the National System of Researchers of CONACYT.  He received his Ph.D. from the National Autonomous University of Mexico.  He can be reached at [email protected] and followed on Twitter @OMCovarrubias

 

(10 votes, average: 5.00 out of 5)
Loading...