Philosopher Carissa Véliz on AI's Ethical Dilemmas in Predictions

In 2020, Carissa Véliz, a Hispanic-Mexican philosopher in her thirties, made a significant impact on the global discourse concerning technology's influence on privacy with her book Privacy is Power (Destino). In this authoritative essay, Véliz addresses the unacceptable nature of surveillance capitalism, highlighting its intrusive effects on personal privacy. Now, six years later, the Oxford University philosophy professor has released a new book titled Prophecy, which will be available in Spain from June via Debate.

In Prophecy, Véliz offers a compelling examination of how statistics and predictions—tools extensively utilized by artificial intelligence (AI)—serve as instruments of power that fundamentally shape our world. She asserts, “Predictions are often orders disguised as a search for knowledge. The entire data economy has emerged because we want to predict; otherwise, we would not waste time and energy on it,” she remarked during an interview in Madrid.

The Allure of Predictions

Question. What is attractive about prophecies?

Answer. Recent years have seen the rise of a divination culture linked to AI. Machine learning fosters a probabilistic mindset alongside the emergence of prediction markets, presenting forecasts as factual. This development carries profound ethical implications.

The Power of Predictions

Q. You suggest that predictions can influence reality, compelling events to align with them.

R. We often misunderstand predictions. My hypothesis is that there is a linguistic illusion; predictions sound definitive, like factual descriptions. However, upon philosophical analysis, we realize they are not. Specifically, human-related predictions affect our expectations, which subsequently shape our reality. They possess a compelling power.

Historical Context of Probability

Q. You assert that although mathematics has ancient roots, the field of probabilities is relatively recent.

R. Indeed. The Greeks were mathematically advanced, yet did not develop a concept of probability. A possible explanation is their belief in fate and divine decision-making, which makes mathematical probability seem unnecessary.

Censuses and Colonialism

Q. You draw parallels between the advent of probability mathematics and censuses as tools of colonialism.

R. Understanding the origins of population statistics is crucial. Francis Galton played a pivotal role in this development. Remarkably, the normal curve was discovered through disparate paths: gaming and the study of stars. Applying statistical tools to social issues, particularly in normative terms, imposed standards of normality on populations. Thus, nonconformity was often labeled as deviance, correlating back to controlling populations, especially marginalized ones.

Social Control and Predictions

Q. Why are statistics critical in terms of social control?

R. Predictions that masquerade as facts create a narrative that can effectively dictate desired outcomes. When people accept these forecasts as fact, they inadvertently conform to the envisioned future.

The Dangers of Categorization

Q. You argue that categorizing individuals erodes personal identity, with AI exacerbating this issue.

R. Correct. Reducing individuals to mere numbers dehumanizes them. In pre-statistical societies, professions were much more fluid. However, once rigid categories were established, those who didn't fit began to experience negative repercussions. This oversimplification enhances statistical precision at the expense of personal nuance, suggesting that statistics are never neutral.

The Role of Bureaucracy and Trust

Q. How did this bureaucratic reliance on numbers evolve?

R. We trust numbers over people, forgetting that humans create these figures. After the decline of divinely justified authority, bureaucrats felt the need to justify their value through numbers, leading to a reliance on impersonal systems that diminish accountability. When errors occur, individuals evade responsibility by attributing faults to these mechanical processes.

The Misleading Nature of Numeric Predictions

Q. Why do statements backed by numbers seem more valid?

R. In academia, presenting a numerical prediction can lead to fame, irrespective of its origin. Often, these figures obscure the matter at hand, disguising fabricated numbers as credible data.

AI's Underlying Risks

Q. Should we view AI as a fragile construct?

R. Absolutely. The illusion of predictability renders us blind to the inherent systemic risks generated by AI, which can't simply be quantified.

A Proactive Approach to Prediction

Q. How can we address the challenges posed by predictions?

R. We should utilize predictions judiciously. While acknowledging their utility—for instance, weather forecasts—we must differentiate between what can and cannot be predicted. Instead of obsessing over uncertain predictions, we should focus on resilience-building measures, such as enhancing ventilation systems in anticipation of future pandemics.

Evaluating the Value of Predictions

Q. Do predictions reflect the world or our knowledge of it?

R. Varied predictions signify a lack of knowledge. Expertise in a field does not equate to expertise in predicting its future; awareness of uncertainties and the unpredictable nature of the future is essential.

The Impact of Utilitarianism

Q. You critique utilitarianism and effective altruism as bureaucratic mechanisms.

R. I recently revisited Charles Dickens, who presented valid criticisms of utilitarians. Their cost-benefit approach has profoundly shaped our moral reasoning. Effective altruists, who argue that extreme wealth is justifiable if it leads to greater charity, have quieted since high-profile controversies but will likely resurface, continuing to influence societal norms.