Axiology

Theory: Axiology | Template: The Deep Dive | Words: 1,841

# Axiology: Values Drive All Research Choices

Many of us in professional life, especially in research and development, have been taught that "objective" means "value-free." We strive to remove our personal beliefs, to let the data speak for itself, and to build systems that are neutral. We spend a lot of time discussing ontology – what is real – and epistemology – how we know it. But there’s a crucial, often-skipped philosophical pillar that underpins both: axiology. This is the study of values, and it reveals that every single research or design decision is, at its core, a value decision.

The Popular Version

The common understanding of values in research often boils down to managing bias. We acknowledge that personal beliefs, cultural backgrounds, or even the desire for a certain outcome can influence a study. The popular solution is to "control for" these biases, to design experiments that minimize their impact, or to remain strictly neutral.

This perspective suggests that if we just gather enough data, use rigorous statistical methods, and apply logic, we can arrive at a purely objective truth. In the world of AI, this idea often manifests as the belief that an algorithm is objective simply because it processes vast amounts of data. The data, we assume, is neutral. The math, we assume, is impartial. Therefore, the outcome must be objective, untainted by human values. It’s a compelling narrative: data as the ultimate arbiter, free from human fallibility.

What the Original Actually Says

The foundational thinkers in axiology offer a far richer and more complex view than simply "personal bias." They argue that values are not just subjective preferences to be eliminated; they are fundamental structures that shape our perception, judgment, and interaction with the world.

Robert Hartman, in his seminal work The Structure of Value (Hartman, 1967), moved beyond the idea of values as mere feelings. He proposed a formal axiology, a structured way to understand value through three distinct dimensions: systemic, extrinsic, and intrinsic. Systemic value relates to logical consistency, like following rules in a game. Extrinsic value is about empirical usefulness, how well something serves a purpose. Intrinsic value, the highest form, recognizes something's unique individuality and inherent worth. Hartman’s work shows that valuing isn’t just a fuzzy feeling; it’s a structured way of apprehending reality.

Max Scheler, another pivotal figure, challenged the idea that ethics could be derived purely from reason (Scheler, 1913/1973). He argued that values are apprehended through feeling, not just logic. Scheler proposed a hierarchy of values – from values of pleasure, to vital values (like health), to spiritual values (like beauty or justice), and finally to values of the holy. For Scheler, we don't invent values; we discover them through our emotional and intuitive experiences. They exist independently of our preferences and guide our moral compass.

Then there's John Dewey, a pragmatist who connected valuation directly to practical judgment and problem-solving (Dewey, 1939). For Dewey, values aren't fixed, abstract ideals. Instead, they emerge from our active engagement with the world as we try to improve our conditions and solve problems. When we decide something is "good" or "better," it's usually in relation to a specific context and a desired outcome. Values, in this view, are instrumental – they guide our actions and help us navigate complex situations. They are not static but evolve as we learn and adapt.

These thinkers, each in their own way, reveal that values are deeply embedded in how we perceive, understand, and act in the world. They are not merely an afterthought or a personal quirk to be controlled, but a fundamental lens through which we interpret reality.

What Changed Since

Axiology has continued to evolve, moving from philosophical foundations to empirical investigation and practical application in technology. This evolution shows that values are not just theoretical constructs but have measurable, universal, and profound impacts.

Shalom Schwartz's extensive cross-cultural research provided empirical evidence for the universality of certain human values (Schwartz, 1992). He identified ten basic values recognized across diverse cultures, including self-direction, security, benevolence, and universalism. This work demonstrates that while cultural expressions differ, a core set of values resonates across humanity. It suggests that values are not entirely arbitrary but reflect shared human needs and experiences.

Elizabeth Anderson further refined our understanding, arguing that values aren't just subjective preferences but are grounded in objective reasons and social practices (Anderson, 2004). This means we can rationally deliberate about values, discuss their merits, and collectively shape them through democratic participation. It moves us away from the idea that "my values are my values, and yours are yours, end of discussion" towards a framework where values can be critically examined and justified within a community.

The rise of information technology and artificial intelligence brought axiology into sharp focus. Luciano Floridi introduced Information Ethics, a field specifically concerned with the ethical impact of information and communication technologies (Floridi, 2013). His work highlights that values like privacy, security, and access to information are not optional add-ons but are fundamental to the design and deployment of digital systems.

This perspective is echoed by researchers like Friedman and Hendry, who explicitly state that value judgments are inherent in the design of computer systems, from the initial problem definition to the final implementation (Friedman & Hendry, 2019). They argue that designers must be acutely aware of the values they are embedding. For example, when an AI system is built, choices are made about what data to collect, what outcomes to prioritize, and how to define success. Each of these choices is a value judgment.

This awareness of embedded values has become a global concern. A 2022 report by the European Commission revealed that an overwhelming 92% of Europeans want AI systems to be developed and used in a way that respects human rights and fundamental values (European Commission, 2022). This isn't just a philosophical debate; it's a societal demand. Similarly, a 2022 survey by the Pew Research Center found that 60% of Americans believe that algorithms should be subject to government regulation (Pew Research Center, 2022). This public sentiment underscores the growing recognition that AI systems, far from being value-neutral, embody and propagate values that profoundly impact society. The call for regulation is a call for explicit value alignment.

As a result, axiology is now central to the development of responsible AI. B.C. Stahl’s work on responsible AI highlights the importance of ethical frameworks and guidelines to ensure AI systems align with human values and societal norms (Stahl, 2021). We see this pattern globally: an increasing recognition that neglecting values in the design phase leads to significant problems downstream.

The Modern Application

So, what does this mean for AI-driven learning and other intelligent systems today? It means we must abandon the myth of value-free research and design. Every decision we make, from the data we collect to the metrics we optimize, is a statement of what we value.

Consider an adaptive learning platform. If we measure "engagement" primarily by "time-on-task," we are implicitly valuing duration over other forms of interaction. If we optimize for "test scores," we prioritize recall over critical thinking or creativity. These aren't just technical choices; they are axiological ones. Neither is inherently "wrong," but they are choices, reflecting what we believe matters most in learning.

The real-world consequences of unexamined values in AI are stark. ProPublica's 2016 investigation into the COMPAS risk assessment algorithm used in the US criminal justice system exposed how it was biased against African Americans (ProPublica, 2016). The algorithm, designed to predict recidivism, predicted higher rates for Black defendants even when controlling for prior criminal history. This wasn't a technical glitch; it was a consequence of embedded values and data choices that perpetuated existing societal inequalities. This aligns with a 2021 study in Nature Machine Intelligence, which found that AI systems trained on biased data can indeed perpetuate and amplify existing social inequalities, leading to discriminatory outcomes (Nature Machine Intelligence, 2021).

Even major tech companies grapple with these ethical dilemmas. In 2018, Google faced significant internal and external criticism over Project Maven, a Pentagon project using AI to analyze drone footage (Google, 2018). Employees protested, citing ethical concerns about the potential use of AI in lethal autonomous weapons. Google ultimately decided not to renew its contract, demonstrating how corporate values can influence business decisions. The NHS, during the COVID-19 pandemic, developed explicit ethical frameworks to guide resource allocation, showing a clear, values-driven approach in a crisis (NHS, 2020).

The cost of ignoring axiology is not just ethical; it's practical. A 2023 report by the World Economic Forum found that a staggering 85% of AI projects fail to deliver on their intended business outcomes, often due to a lack of attention to ethical and societal considerations (World Economic Forum, 2023). This isn't just about "doing good"; it's about building effective, sustainable, and trusted systems.

The Reference Guide

Understanding axiology moves us beyond simply acknowledging "bias" to actively shaping our values. Here are key principles to guide our approach to research and AI development:

• Values are Foundational: Values are not subjective preferences to be eliminated. They are fundamental structures that shape how we perceive and judge the world, influencing our ontology (what is real) and epistemology (how we know).
• Structured and Apprehended: Values can be understood in structured ways (Hartman, 1967) and are apprehended through feeling and intuition, not just pure reason (Scheler, 1913/1973).
• Contextual and Evolving: Values emerge from our interactions with the world and our attempts to solve problems, adapting as conditions change (Dewey, 1939).
• Universal Patterns Exist: While culturally expressed differently, certain core human values are recognized across societies (Schwartz, 1992).
• Values are Justifiable: Values can be grounded in objective reasons and social practices, allowing for rational deliberation and collective shaping (Anderson, 2004).
• Inherent in Design: Values are inherently embedded in the design of all systems, especially AI. Every design choice, from data selection to optimization goals, is a value judgment (Friedman & Hendry, 2019).
• Transparency is Key: Pretending research or AI is value-free doesn't remove values; it merely hides them, making them harder to scrutinize and leading to unintended, often harmful, consequences.
• Responsible AI Demands Axiology: Explicitly naming and aligning AI systems with human values and societal norms is crucial for ethical, effective, and trustworthy technology (Stahl, 2021).

The Challenge

The next time someone declares a research finding "objective" or an algorithm "neutral" because "it's just data," challenge them to dig deeper. Ask them what values are implicitly guiding that objectivity. We know from a 2020 study by AlgorithmWatch that 70% of algorithmic systems used in Europe lack transparency regarding their decision-making processes (AlgorithmWatch Report, 2020). This lack of transparency directly obscures the values at play.

True rigor isn't about pretending values don't exist. It's about recognizing them, naming them, and engaging with them transparently. It's about moving from axiological illiteracy to intentional, value-driven design.

Whose values are embedded in your AI system?