When more is not better: Rethinking rationality in the age of AI

By Richard Sebaggala (PhD)

Economists love simple assumptions, and one of the most enduring is the idea that more is better, or the non-satiation principle. More income, more production, more consumption: in our economics textbooks, a rational actor never rejects an additional unit of utility. By and large, this principle has proven to be reliable. Who would turn down more wealth, food or opportunity? However, there are exceptions. In monogamous marriages, “more” is rarely better and certainly not rational. Such humorous caveats aside, this assumption has informed much of our understanding of economic behaviour.

 

Economists refer to this principle as the monotonicity assumption, i.e. the idea that consumers always prefer more of a good over less. As Shon (2008) explains, monotonicity underpins key findings of microeconomics: utility maximisation takes individuals to the limit of their budget, and indifference curves cannot intersect. Even Gary Becker, who argued that monotonicity need not be explicitly assumed, concluded that rational agents behave as if “more is better” because they adjust their labour and consumption up to that point. In short, the discipline has long assumed that “more” is a safe rule of thumb for rational decision-making.

 

Artificial intelligence poses a challenge to this axiom. While most people recognise its potential, many are quick to emphasise the risks of overreliance, focusing on the negative impacts and overlooking the benefits that come from deeper engagement. My own experience is different. The more I use AI, the better I get at applying it to complex problems that once seemed unsolvable. It sharpens my thinking, increases my productivity and reveals patterns that were previously difficult to recognise. However, the critics are often louder. A recent essay in the Harvard Crimson warned that students use ChatGPT in ways that weaken human relationships: they look for recipes there instead of calling their mothers, they consult ChatGPT to complete assignments instead of going to office hours, and they even lean on ChatGPT to find companionship. For the author, any additional use of AI diminishes the richness of human interaction.

 

This view highlights a paradox. A technology that clearly creates abundance also creates hesitation. Economics offers a few explanations. One of them is diminishing marginal utility. The first experience with AI can be liberating as it saves time and provides new insights. However, with repeated use, there is a risk that the benefits will diminish if users accept the results uncritically. Another problem is that of external effects. For an individual, using ChatGPT for a task seems rational- faster and more efficient. However, if every student bypasses discussions with fellow students or avoids professors’ office hours, the community loses the opportunity for dialogue and deeper learning. The private benefit comes with a public price.

 

There is also the nature of the goods that are displaced. Economists often assume that goods are interchangeable, but AI shows the limits of this logic. It can reproduce an explanation or a recipe, but it cannot replace friendship, mentorship or the warmth of a shared conversation. These are relational goods whose value depends on their human origin. Finally, there is the issue of bounded rationality. Humans strive for more than efficiency; they seek belonging, trust and reflection. If students accept AI’s answers unquestioningly, what seems efficient in the short term undermines their judgement in the long term.

 

It is important to recognise these concerns, but it is equally important not to let them obscure the other side of the story. My own practise shows that the regular, deliberate use of AI does not lead to dependency, but to competence. The more I engage with it, the better I get at formulating questions, interpreting results and applying them to real-world problems. The time previously spent on routine work is freed up for thinking in higher dimensions. In this sense, the increased use does not make me less thoughtful but allows me to focus my thoughts where they are most important. So, the paradox is not that more AI is harmful. The problem is unthinking use, which can crowd out the relational and cognitive goods we value. The solution lies in balance: using AI sufficiently to build capabilities while protecting spaces for human relationships and critical engagement.

 

The implications are far-reaching. If AI undermines reflection, we weaken human capital. If it suppresses interaction, we weaken social capital. Both are essential for long-term growth and social cohesion. However, if we use AI as a complement rather than a substitute, it can strengthen both. This is important not only at elite universities, but also in African classrooms where I teach. Here, AI could help close resource gaps and expand access to knowledge. But if students only see it as a shortcut, they will miss out on the deeper learning that builds resilience. Used wisely, however, AI can help unlock skills that our education systems have struggled to cultivate.

 

For this reason, I characterise my perspective as pragmatic. I do not ignore the risks, nor do I believe that technology alone guarantees progress. Instead, I recognise both sides: the fears of those who see AI undermining relationships, and the reality that regular, deliberate use will make me better at solving problems. The challenge for economists is to clarify what we mean by rationality. It is no longer enough to say that more is always better. Rationality in the age of AI requires attention to quality, depth and sustainability. We need to measure not only the efficiency of obtaining answers, but also the strength of the human and social capital we obtain in the process.

 

So yes, more is better, until it isn't. The most sensible decision today may be to put the machine aside and reach out to a colleague, a mentor or a friend. And when it's time to return to the machine, do so with sharper questions and clearer judgement. In this way, we can preserve the human while embracing the transformative. That, I believe, is how to think like an economist in the age of AI.