Ron Aboodi

As Artificial Intelligence (AI) keeps advancing, Generation Alpha and future generations are more likely to cope with situations that call for critical thinking by turning to AI and relying on its guidance without sufficient critical thinking. I defend this worry and argue that it calls for educational reforms that would be designed mainly to (a) motivate students to think critically about AI applications and the justifiability of their deployment, as well as (b) cultivate the skills, knowledge, and dispositions that will help them do so. Furthermore, I argue that these educational aims will remain important in the distant future no matter how far AI advances, even merely on outcome-based grounds (i.e., without appealing to the final value of autonomy, or authenticity, or understanding, etc.; or to any educational ideal that dictates the cultivation of critical thinking regardless of its instrumental value). For any “artificial consultant” that might emerge in the future, even with a perfect track record, it is highly improbable that we could ever justifiably rule out or assign negligible probability to the scenario that (a) it will mislead us in certain high-stakes situations, and/or that (b) human critical thinking could help reach better conclusions and prevent significantly bad outcomes.

A teacher controls the release of materials in attempt to get students to appreciate the appeal of a popular yet wrongheaded argument before exposing them to its shortcomings. An instructor uses body language, tone of voice, and images in a Power-Point presentation that appeal to non-deliberative mechanisms in order to influence the students to pay more attention, maintain their focus, or to remember the content better. How do we draw the line between such innocuous educational practices and problematic manipulation, such as deterring students from questioning certain views by instilling fear or shame? I help answering this question by illuminating one significant danger, which hasn’t been accurately identified in the relevant literature: the danger that manipulations in education—even when they’re non-deceptive and aim for the good of the students—will hinder the development or fulfillment of the students’ deliberative projects.

Jennifer Rose Carr’s (2020) article “Normative Uncertainty Without Theories” proposes a method to maximize expected value under normative uncertainty without Intertheoretic Value Comparison (hereafter IVC). Carr argues that this method avoids IVC because it avoids theories: the agent’s credence is distributed among normative hypotheses of a particular type, which don’t constitute theories. However, I argue that Carr’s method doesn’t avoid or help to solve what I consider as the justificatory problem of IVC, which isn’t specific to comparing theories as such. This threatens the implementability of Carr’s method. Fortunately, I also show how Carr’s method can nevertheless be implemented. I identify a type of epistemic states where the justificatory problem of IVC is not a necessary obstacle to maximizing expected value. In such states, the uncertainty stems from indecisive normative intuitions, and the agent justifiably constructs each normative hypothesis on the basis of a consistent subset of her intuitions by reference to the same unit of value. This part of my argument complements not only Carr’s (2020) argument, but also some moderate defenses of IVC. The combination of Carr’s paper and mine helps to illuminate the conditions for maximizing expected value under normative uncertainty without unjustified value comparison.

How should deontological theories that prohibit actions of type K — such as intentionally killing an innocent person — deal with cases of uncertainty as to whether a particular action is of type K? Frank Jackson and Michael Smith, who raise this problem in their paper "Absolutist Moral Theories and Uncertainty" (2006), focus on a case where a skier is about to cause the death of ten innocent people — we don’t know for sure whether on purpose or not — by causing an avalanche; and we can only save the people by shooting the skier. One possible deontological attitude towards such uncertainty is what Jackson and Smith call the threshold view, according to which whether or not the deontological constraint applies depends on our degree of (justified) certainty meets a given threshold. Jackson and Smith argue against the threshold view that it leads to implausible paradoxical moral dilemmas in a special kind of case. In this response, we show that the threshold view can avoid these implausible moral dilemmas, as long as the relevant deontological constraint is grounded in individualistic patient-based considerations, such as what an individual person is entitled to object to.

De dicto moral motivation is typically characterized by the agent’s conceiving of her goal in thin normative terms such as to do what is right. I argue that lacking an effective de dicto moral motivation would put the agent in a bad position for responding in the morally-best manner in a certain type of situations. Two central features of the relevant type of situations are the appropriateness of the agent’s uncertainty concerning her underived moral values, and the practical, moral importance of resolving this uncertainty. I argue that in some situations that are marked by these two features the most virtuous response is deciding to conduct a deep moral inquiry for a de dicto moral purpose. In such situations lacking an effective de dicto moral motivation would amount to a moral shortcoming. I show the implications for Michael Smith’s argument against Motivational Judgment Externalism and for Brian Weatherson’s argument against avoiding moral recklessness: both arguments rely on a depreciating view of de dicto moral motivation, and both fail; or so I argue.

I argue that there are (only) two contingent factors that can render an instantiation of de dicto moral motivation—which is typically characterized by the agent's conceiving of her goal in moral terms such as doing what's right—less virtuous than some alternative motivation that would lead to the same (right) action: (1) the circumstances are such that it would be more virtuous to be moved directly by certain non-deliberative dispositions (such as an emotional attachment to one's spouse); or (2) the circumstances are such that de dicto moral motivation has practical disadvantages (such as generating unnecessary moral reflections that waste precious time).