In the previous blog post Levinas’ ethics of infinite responsibility toward the Other was introduced as a lens for evaluating the role of artificial intelligence in drafting recommendations. The blog post concerned a specific use case: translating adolescent students’ belief-speaking answers into recommendations. In this blog post I extend my inquiry to a more general use case: responses to Foreign Information Manipulation and Interference (FIMI) to assess whether AI should or should not be allowed to draft recommendations in this domain.
Levinas’ approach is a solid fit with the adolescent students’ use case, as I will show below when I’ll be considering the aspects of of contextual viability, reciprocity potential, scale of impact, moral intuition, self-consistency, and existential risk. The fit renders Levinas’ ethical frame applicable and, by implication, is the reason why AI may not be used to draft recommendations. Next I will assess whether Levinas’ work is also a good fit for those in the FIMI domain, and what the implications of this assessment are for the role of AI in drafting recommendations.
Contextual viability
Contextual viability concerns assessing the practical feasibility of applying Levinas’ approach. If the context allows for radical generosity without catastrophic loss (e.g., abundant resources, low risk to self), leaning toward infinite duty might be viable. Adolescents are often vulnerable, dependent, and in a formative stage. Offering them a voice aligns with Levinas’s focus on the Other’s uniqueness and vulnerability, treating their needs as a moral summons. The cost to the self (adults, society) is typically manageable — listening and adapting to them almost never threaten survival.
Reciprocity potential
Reciprocity potential equals evaluating whether the Other is likely to respond in kind or exploit the gesture. In the case of adolescents, reciprocity is irrelevant to the ethical act since supporting adolescents is a moral and legal obligation, even though that obligation diminishes with the age of adolescents. It seems fair to accept the gravitas of adolescent voices as an unconditional duty regardless of the Other’s response.
Scale of impact
It is important to consider the scope of who is affected by an approach — individuals, groups, or systems — and its long-term consequences. According to Levinas, if the focus is on the immediate, singular Other (the face before you), infinite responsibility might dominate. In the adolescent student use case, we meet a large collection of singular Others – they all voice their answers independently – but we do not meet them in the flesh; the meeting is mediated. Nevertheless, the upside regarding impact is similar: by meeting adolescent input with ethical responsibility, we will probably foster their long-term trust and psychosocial integration, benefiting both the individual and the collective over time.
Moral intuition
We need to weigh intuitive ethical impulses against calculated outcomes. According to Levinas, if a visceral, pre-rational pull toward the Other’s suffering drives us (Levinas’s "face" as a moral summons), we might lean toward duty. In the case of adolescent voices, this intuitive ethical pull to nurture and empower adolescents exists, seeing their needs as a call we’re obligated to answer, even if it requires sacrifice (e.g., time, resources, or shifting priorities).
Self-consistency
We need to reflect on whether the choice for Levinas’ approach aligns with your core values or sense of self. If our identity is tied to being a moral exemplar or transcending egoism, we might gravitate toward infinite responsibility. This is a crucial issue: do we, as caregivers and as a society, value compassion? If yes, then this aligns with an identity of care and openness to the Other’s perspective. If not, if we declare that care and openness are no core values that we cherish, then Levinas’ approach will feel like alien. As established earlier, we do proclaim to hold these values dear with regard to our adolescents, so we should be open to their voices.
Existential risk
As we saw above listening and adapting to adolescent voices almost never threaten our survival. In my opinion rather the opposite seems true: by not listening and adapting we (further) enable the current frequently occurring adolescent condition of partial dislocation – a condition of psychosocial integration within a small group while experiencing dislocation in society at large. As was written in blog post 31, this condition enables more adolescent openness to radical solutions, from substance abuse to radical ideologies and religions, violence, and mental health problems.
Use case FIMI
While Levinas’ approach is a good fit for the use case involving belief-speaking adolescents, and as a result is a decisive argument against using AI to be used for drafting recommendations, it may very well be different for other use cases. Without the applicability of Levinas’ frame, the option reemerges to use AI to draft recommendations. Below I consider the use case of dealing with FIMI. Does Levinas’ approach fit this use case as well, and can we thus conclude that AI may not be used to draft recommendations on how to deal with FIMI?
Target groups
When dealing with FIMI, three target groups are relevant: the demand-side (at least a part of the domestic citizens and potentially all of them), the supply-side (those creating and disseminating FIMI: foreign actors and their proxies), and the enablers (those who make the dissemination of FIMI possible).
Levinas’ ethical approach is only relevant when dealing with individuals for whom we want to or should take responsibility. This excludes any interactions with organisations and bots from consideration. They are beyond the realm of Levinas’ thinking. Just like AI they are no “bodies among bodies”.
The same goes for anybody who does not or cannot speak and act freely as an individual. They are not capable of escaping what Levinas calls “totality”, a greater whole that defines them while they define us. What invokes our responsibility is a radical encounter with the Other, which is only possibly beyond the real of “totality”. While such encounters might occur with those fulfilling roles that define them, they can only occur in their role of private, intimate individuals, not in their defined, total roles.
For all these categories Levinas is a no fit; for them a different ethical framework is needed. The fact that Levinas is off the table means that the option of AI drafting recommendations concerning these categories is undecided yet.
Potential good fits with Levinas are the target groups domestic and foreign individuals; they are individuals for whom we potentially might or should have an ethical concern.
Domestic citizens
The first target group to consider are domestic citizens. How does this group fit in Levinas’ frame?
First set of aspects
The first aspects to consider are contextual viability and existential risk. Is it practically viable to give a voice to every citizen, given the costs? The answer so far by representative democracies is “yes”, if with restrictions. One of these restrictions is that we use mediators, such as elections and referenda, for citizens to regularly use their voice. Another restriction is that in these elections and referenda certain types of voices are excluded. These exclusions are incompatible with Levinas’ ethics, under the precondition that the Other is open to potentially life-changing face-to-face meetings. Lifting the restrictions might constitute a risk but a risk that is legitimized by our moral obligation to meet others.
Theoretically, it would be possible to introduce “participation hours” for all citizens to enable less mediated expressions of citizens’ opinions, just like my wife and I propose for adolescents. Practically, it will be much more difficult to implement because most citizens, unlike adolescents, do not meet according to the regulations of the administration. Whereas “participation hours” can be imposed on the school curriculum so that all adolescents enjoy these at the same time under the same didactical rules, this does not hold good for adults.
Second set of aspects
Then, there are the aspect of reciprocity potential, scale of impact, moral intuition, and self-consistency. They tie to responsiveness, an important pillar of liberal democracies. Responsiveness means that the government forms and implements policies that citizens want. It is both a legitimation of democracy and a desirable quality of performance. Diminished responsiveness leads to what Yasha Mounk calls “undemocratic liberalism”: “elites are taking hold of the political system and making it increasingly unresponsive: the powerful are less and less willing to cede to the views of the people.” So, facing the voice of domestic citizens is self-consistent.
The fit is reinforced by the expected scale of impact: enhanced citizen resilience. In project SAUFEX we’ve defined resilience in relation to dealing FIMI as a two-way street (see blog post four): on the one hand it means providing safety for domestic citizens by means of vigilance against foreign activities that aim to promote polarisation, alienation, learned helplessness, relativism, and nihilism, and that stress threats to our physical and psychological health, while at the same time supporting citizens’ experience of belonging and autonomy. This implies that the reciprocity potential equals less domestic citizens turning away from liberal democracy as a result of (partial) dislocation, while the scale of the impact is anyone endangered by dislocation. Both expected outcomes show the weight of its moral intuition.
Non-domestic citizens
When considering non-domestic citizens — particularly those on the supply side of FIMI — Levinas’ ethical framework appears ill-suited. Foreign actors relevant for FIMI are those who create and disseminate FIMI with intent to harm, or at minimum, accept harm as a byproduct of their pursuits (e.g., financial gain). This stance conflicts with Levinas’ emphasis on openness to the Other and the possibility of escaping "totality" through a radical, face-to-face encounter. Unlike domestic citizens or adolescents, for whom we bear a moral or legal duty to respond, non-domestic actors engaged in FIMI show no such reciprocal vulnerability. Levinas’ call for infinite responsibility falters here.
Conclusion
Levinas’ ethical framework, with its emphasis on infinite responsibility toward the singular, vulnerable Other, provides a compelling lens for determining when AI should not be used to draft recommendations. A categorical ban on AI-generated recommendations applies in cases where the Other is an individual for whom we bear, or ought to bear, a direct moral responsibility — marked by contextual viability, moral intuition, and self-consistency with values of care and openness. For adolescent students, this holds true: their vulnerability, dependence, and formative stage summon an unconditional duty to hear their voices, unmediated by AI, fostering trust and psychosocial integration. Similarly, with domestic citizens facing FIMI, Levinas’ ethics aligns with our obligation to enhance resilience and responsiveness. In both cases, a ban on AI to draft recommendations is justified.
The ban does not extend to cases where Levinas’ approach does not apply — namely, with non-domestic citizens on the supply side of FIMI , domestic citizens tied up in their total roles, and entities like organizations and bots. These actors, lacking the reciprocal vulnerability or individual "face" central to Levinas’ thought, fall outside the scope of infinite responsibility. Here, another ethical frame is needed, a frame that possibly does not exclude the use of AI to draft recommendations.
