This blog post delves deeply into how AI decisions impact human autonomy and happiness.
Advances in science and technology have enabled diverse attempts to make human life more convenient and happier. Among these, the scope of artificial intelligence’s application continues to expand daily. In the future, individuals will actively provide their information to receive better services, and as a result, artificial intelligence will gain access to sensitive information such as personal genetic test results, biometric data, and medical records.
As the scope of information collected by AI broadens, it ultimately becomes possible to access an individual’s entire private life. Furthermore, there is a possibility that AI systems could emerge that understand the user better than the user understands themselves. If this happens, humans will gradually entrust important decisions to AI. This stems from the expectation that AI can analyze a user’s tendencies and preferences to derive the most satisfying outcome. Ultimately, the situation where the decision-maker shifts from humans to AI could become a reality.
However, at this point, there is an important question we must consider. Human life is a continuous series of choices. An individual’s life can change direction based on the decisions they make, and the very nature of their life can transform entirely. The act of ‘decision-making’ holds profound significance in life. So, can decisions made by data-driven AI truly make humans happier? Even if the decision isn’t one we agonized over ourselves, if an AI system presents a path with a high probability of satisfaction based on accumulated data, and life proceeds accordingly, can we be certain that such a life is better than one lived through autonomous decision-making?
On the surface, it might always seem to lead to satisfying decisions. However, it is questionable whether that momentary satisfaction, and decisions made by AI instead of through one’s own deliberation, can truly make humans happier. The reasons are as follows.
First, AI analyzes past choices and their resulting satisfaction levels to make subsequent decisions. Consequently, users may be continuously guided only toward decisions they’ve made before, especially those with high satisfaction. This can fundamentally block the possibility of choosing new directions. Yet, in reality, new choices different from the past can bring greater satisfaction, and sometimes an unsatisfying decision can unexpectedly lead to positive outcomes.
The reason AI struggles to make such novel decisions is the absence of satisfaction data for those experiences. Consequently, the system prioritizes previously validated choices and is highly likely to exclude new attempts. Yet in reality, an activity never tried before can sometimes bring significant satisfaction.
Of course, counterarguments exist. If AI possesses sufficient user data, it could analyze a user’s inclinations or aptitudes to make appropriate suggestions, even for new activities the user hasn’t experienced. For example, if a user shows active tendencies, AI might suggest, “People with similar inclinations enjoy swimming. Why not give it a try?”
However, this too is merely a suggestion based entirely on the satisfaction data of ‘other people’. It’s impossible to know whether the specific user will actually enjoy swimming, and even if they do, predicting the level of their satisfaction is difficult. Ultimately, AI recommends activities within a predictable range and based on past data, where satisfaction is relatively assured.
For example, if AI analyzes that the user would gain 70% satisfaction from swimming and 80% from soccer, the user will likely stick with soccer, which they’ve tried before. Consequently, the user may permanently miss the chance to try swimming and might never discover the possibility that swimming could become their new hobby or specialty.
This solidifies past choice patterns, ultimately leading to closed-mindedness.
Moreover, satisfaction with a specific activity can gradually decrease the more it is repeated. What initially felt new and enjoyable becomes familiar with repetition, increasing the likelihood of boredom. This is similar to how a variety show with high ratings loses popularity after repeatedly using a similar format. Trying diverse activities can actually provide newer forms of satisfaction. Such unpredictable satisfaction cannot be found in the existing data possessed by artificial intelligence. Therefore, the possibility always exists that making one’s own decisions, rather than following AI decisions, will bring greater satisfaction.
Furthermore, AI always makes decisions based on existing data that are likely to yield high satisfaction, so it has no reason to intentionally make unsatisfying choices. However, in human life, unsatisfying decisions can sometimes lead to entirely unexpected new possibilities.
Consider an example. When choosing a romantic partner from among people you already know, AI could recommend someone with high satisfaction based on existing data. Yet, through a chance mistake or unexpected circumstance, you might meet a stranger who turns out to be a surprisingly better match. Such cases where an unsatisfactory decision leads to greater satisfaction are difficult for AI to predict.
Of course, one could counter that AI can predict the outcomes of each decision and that satisfaction analysis can account for all impacts resulting from those outcomes. However, encounters between humans, especially the formation of relationships with people met by chance for the first time, are statistically and structurally difficult to predict. Such encounters are based purely on randomness and coincidence, and simultaneously involve complexly intertwined decisions made by multiple people, inevitably leading to significantly lower prediction accuracy for AI.
Most crucially, if humans merely follow AI decisions, they lose opportunities to experience mistakes or failures. AI intervention drastically reduces the likelihood of error or failure, thereby depriving humans of the chance to learn and grow from setbacks.
Of course, counterarguments exist that some things can be learned without failure, and growth is possible without experiencing setbacks. However, failure holds meaning beyond mere learning. It can sometimes deliver a psychological shock to humans, compelling them to redefine their life’s direction and values. Through such painful periods of reflection, humans come to realize their intrinsic worth and gain profound insights into life. The experience of failure and frustration makes humans stronger, and that is precisely the driving force for growth that success cannot provide.
As technology advances further, artificial intelligence systems deeply intervening in human life based on vast personal data will indeed emerge. The more the introduction of algorithms for ‘better,’ ‘more convenient,‘ and ‘happier’ lives becomes reality, the more we have reached a point where we must seriously consider whether such technology can truly lead humans to a better life.
In this context, I believe we should value the unique worth that humans can create through their own thinking and as human beings themselves, rather than placing blind trust in AI’s judgments. Human choices, mistakes, deliberations, and failures carry meaning beyond mere variables; they are the most essential elements constituting a truly human life.
