Responsible Black Boxes: How Virtue Ethics Can Bridge the Responsibility Gap in AI (Palgrave Handbook Ch 22)

Hasse J. Hällström; Steven S. Gouveia · 2026 · Palgrave Handbook on the Ethics of AI, ch. 22, pp. 335-355 interlocutor high priority coded

Main argument

Thesis: the responsibility gap created by black-box opacity cannot be bridged by XAI (the prevailing assumption that interpretability enables blame assignment fails: opacity is partly irreducible, and interpretability doesn't itself create accountable conduct) nor by principlism (codes struggle under opacity); Aristotelian VIRTUE ETHICS - stable character traits and phronesis located in ENGINEERING ORGANISATIONS - offers 'a more robust path to Responsible AI, even if the systems themselves remain (partly or entirely) unexplainable': 'By highlighting moral dispositions that guide engineering organisations, VE ensures that accountability does not hinge on model transparency. Instead, conscientious engineering organisations demonstrate honesty, responsibility, and courage to address emerging harm, correct biases, and openly acknowledge uncertainties.' Feasibility honestly doubted: market forces reward rapid deployment over deliberation; exemplars with both ethical wisdom and technical prowess are rare - so the chapter also addresses how to incorporate VE into AI development culture.

Why it matters here

The virtue-ethics bridge over the responsibility gap that DECOUPLES responsibility from explainability: organisational virtue (honesty, responsibility, courage in engineering organisations) grounds accountability even for systems that remain unexplainable - 'XAI is not necessary for responsibility, provided that engineers act virtuously.' The direct rival to Kästner's MI-based (epistemic-access) allocation - the dissertation must adjudicate between them.

Reading notes

Close read (21pp; Helsinki industry + Porto - the handbook editor's own chapter). Structure: black-box problem; XAI-based and principlist solutions fail under opacity; Aristotelian virtues in ENGINEERING ORGANISATIONS as the responsibility ground; feasibility worries (market incentives reward speed; moral exemplars rare).

Hällström, H. J., & Gouveia, S. S. (2026). Responsible Black Boxes: How Virtue Ethics Can Bridge the Responsibility Gap in AI. In S. S. Gouveia (Ed.), The Palgrave Handbook on the Ethics of AI (ch. 22). Palgrave.

Close reading — 3 coded units

#1 · pp. 335 · claim

“we argue that virtue ethics (VE) rooted in Aristotle's account of stable character traits and phronesis (practical wisdom) presents a more robust [path to 'Responsible AI', even if the systems themselves remain (partly or entirely) unexplainable].”

NF-VIRTUE

#2 · pp. 336 · argument

“By highlighting moral dispositions that guide engineering organisations, VE ensures that accountability does not hinge on model transparency. Instead, conscientious engineering organisations demonstrate honesty, responsibility, and courage to address emerging harm, correct biases, and openly acknowledge uncertainties. [...] XAI is not necessary for responsibility, provided that engineers act virtuously.”

RL-DEV RL-EPISTEMIC

#3 · pp. 336 · objection

“However, the feasibility of this virtue-ethical approach is debated. Industry and market forces often reward rapid deployment at the cost of careful moral deliberation. Furthermore, the implementation of VE requires moral exemplars with both ethical wisdom and technical prowess, which can be rare in practice.”

RL-INST

Synthesis-matrix row

Memos (1)

theoretical · unit #2

The Kästner-vs-Hällström adjudication is a ready-made dissertation section: Kästner et al. make responsibility TRACK epistemic access (MI shifts liability to whoever understands); Hällström & Gouveia DECOUPLE them (organisational virtue grounds responsibility regardless of understanding). The dissertation can argue these are complementary at different timescales: virtue governs EX-ANTE conduct under irreducible opacity (choosing to deploy, monitoring, acknowledging uncertainty - Gregg's escalation layer), while difference-making + epistemic access allocates EX-POST liability for specific harms - and the FMTI transparency decline (Stanford HAI) shows why neither alone suffices: virtue without disclosure is unverifiable (organisations can claim conscientiousness - the safety-washing worry, H&D fn25), disclosure without virtue is Akyol's ethical hallucination at the organisational level. Synthesis: virtue supplies the standard of care, difference-making the attribution machinery, institutions the verification - the three-level allocation again.