Artificial intelligence is no longer a distant promise. It is embedded in everyday systems that shape access to credit, employment, healthcare, security, and public services. Increasingly, decisions that affect people’s lives are informed or even determined by algorithms. This shift raises a fundamental question: who actually decides in the age of AI?
At first glance, algorithmic decision making appears efficient and objective. By processing large volumes of data, AI systems promise consistency, speed, and predictive accuracy. Governments and institutions are adopting these systems to improve service delivery, optimize resources, and reduce human error. Yet beneath this promise lies a more complex reality. Algorithms do not eliminate power. They reorganize it.
From Human Judgment to Algorithmic Decision Making
Traditionally, decision making in public policy and institutions has relied on human judgment, guided by rules, experience, and discretion. With the rise of AI, this process is increasingly mediated by models that classify, predict, and recommend.
These systems are used in diverse domains. In finance, algorithms assess creditworthiness. In public administration, they support welfare targeting and risk assessment. In law enforcement, predictive tools are used to anticipate crime patterns. In hiring, automated systems filter candidates based on predefined criteria.
While these applications differ, they share a common feature. Decisions are shaped by statistical patterns derived from historical data. This introduces a critical concern. If the past reflects inequality, algorithmic systems may reproduce and even amplify it.
Research has shown that algorithmic systems can inherit and reinforce bias embedded in data. For example, risk assessment tools in criminal justice have been found to produce disparate outcomes across racial groups, raising questions about fairness and accountability (Angwin et al., 2016). Similarly, studies of algorithmic fairness highlight how technical systems can encode social inequalities in ways that are difficult to detect and challenge (Barocas and Selbst, 2016).
The Question of Power
To understand algorithmic decision making, it is necessary to move beyond technical explanations and examine power.
Algorithms are not autonomous actors. They are designed, trained, and deployed by institutions. The choices made in this process matter. Decisions about which data to use, which variables to prioritize, and which outcomes to optimize are inherently political.
Power operates at multiple levels. It is present in the design of models, in the ownership of data, and in the institutional contexts where systems are applied. Large technology companies and state agencies often control the infrastructure and expertise required to develop and deploy AI systems. This creates asymmetries that shape who has influence over decision making processes.
Moreover, algorithmic systems can obscure power rather than make it visible. Decisions that were once open to scrutiny may now be framed as outputs of neutral systems. This phenomenon, often described as the opacity of algorithms, makes it more difficult to question how and why decisions are made (Pasquale, 2015).
Automation and the Risk of Deference
Another important dynamic is the tendency toward automation bias. When decisions are supported by algorithmic systems, there is a risk that human actors defer to machine outputs, even when those outputs are flawed.
This deference is reinforced by the perceived authority of data and technology. When a decision is presented as data driven, it is often seen as more credible and less contestable. However, this perception can mask underlying assumptions and limitations.
Algorithmic systems are only as reliable as the data and models that underpin them. They are shaped by incomplete information, modeling choices, and contextual constraints. Treating them as infallible can lead to unjust outcomes, particularly for individuals and groups that are poorly represented in data.
Algorithmic Governance and Accountability
The growing reliance on AI in decision making has led to what is often described as algorithmic governance. In this context, rules and decisions are increasingly embedded in technical systems rather than formal institutions alone.
This shift raises important questions about accountability. Who is responsible when an algorithm produces a harmful or unjust outcome? Is it the developer, the institution that deploys the system, or the decision maker who relies on its output?
Current governance frameworks often struggle to address these questions. Algorithmic systems can be complex and difficult to interpret, making it challenging to assign responsibility or provide meaningful explanations to those affected.
Scholars have argued for greater transparency, auditability, and oversight of algorithmic systems to ensure accountability and fairness (Diakopoulos, 2016). However, implementing these principles in practice remains a significant challenge.
Algorithmic Power and the Data Justice Perspective
These issues can be better understood through the lens of data justice.
Algorithmic systems influence who is represented in data, how resources are distributed, and who has control over decision making processes. This aligns with three key dimensions.
Representation concerns which individuals and communities are visible in the data used to train models. When certain groups are underrepresented or misrepresented, the resulting systems may fail to serve them adequately.
Distribution relates to how benefits and burdens are allocated. Algorithmic decisions can determine access to opportunities and services, shaping patterns of inclusion and exclusion.
Governance addresses who controls the systems and sets the rules. The concentration of technical expertise and data resources in a limited number of actors raises concerns about democratic oversight and participation.
Understanding algorithmic power through these dimensions highlights that AI is not just a technical tool. It is a site where questions of justice are negotiated.
Rethinking Decision Making in the Age of AI
The challenge is not to reject AI, but to critically engage with how it is designed and used.
This requires a shift in perspective. Instead of asking whether algorithms are accurate, it is necessary to ask whether they are fair, accountable, and aligned with societal values.
Practical steps include improving data quality and inclusiveness, conducting regular audits of algorithmic systems, and ensuring that human oversight remains meaningful rather than symbolic. It also involves creating mechanisms for affected individuals to contest decisions and seek redress.
Importantly, decision making should not be fully delegated to technical systems. Human judgment, contextual understanding, and ethical reflection remain essential.
Conclusion
In the age of AI, decisions are increasingly shaped by systems that operate beyond immediate human perception. This does not mean that power has disappeared. It has been redistributed and, in some cases, obscured.
The central question is not whether algorithms decide, but how decisions are structured, who has influence over them, and who bears their consequences.
Understanding algorithmic power is therefore essential to ensuring that the expansion of AI contributes to justice rather than deepening inequality.
References
Angwin, J., Larson, J., Mattu, S., and Kirchner, L. (2016). Machine Bias. ProPublica.
Barocas, S., and Selbst, A. (2016). Big Data’s Disparate Impact. California Law Review.
Pasquale, F. (2015). The Black Box Society. Harvard University Press.
Diakopoulos, N. (2016). Accountability in Algorithmic Decision Making. Communications of the ACM.
samivan.id 
Leave a comment