Luciano Floridi (2013), Science and Engineering Ethics 19(3): 727–743 — DOI-10.1007/s11948-012-9413-4
Key takeaways
- Morally loaded actions can come from interactions that are each morally neutral or negligible. A multiagent system can do good or evil that none of its members does. Floridi calls this distributed morality (DM). It covers only what does not reduce to “the sum of (some) human, individual, and already morally-loaded actions” (p. 729).
- The mechanism is a pair of moral thresholds. Possible evils below them fail (the environment is “fault-tolerant”), and possible goods below them stay neutral (it is “inert”). So the ethical task includes designing aggregation of possible goods and fragmentation of possible evils (p. 736).
- “Infraethics” names the first-order framework of moral enablers, such as trust, privacy, transparency and openness, which facilitate good or evil and are “not necessarily morally good in itself” (pp. 738–739).
The paper’s own abstract:
The phenomenon of distributed knowledge is well-known in epistemic logic. In this paper, a similar phenomenon in ethics, somewhat neglected so far, is investigated, namely distributed morality. The article explains the nature of distributed morality, as a feature of moral agency, and explores the implications of its occurrence in advanced information societies. In the course of the analysis, the concept of infraethics is introduced, in order to refer to the ensemble of moral enablers, which, although morally neutral per se, can significantly facilitate or hinder both positive and negative moral behaviours.
Standing in the record
Filed on 2026-10-10 with the paper the owner asked for, Floridi 2025
(LIT-897), which cites it once. It was filed Deferred, unread, because no
open copy could be reached. The owner then supplied the published version,
and NOTE-698 is the close reading of the same day. It placed the work:
Active, for the reasons in the status note.
It extends Floridi & Sanders 2004 (LIT-898): “I introduced the concept of
distributed morality (DM) in (Floridi and Sanders 2004)” (p. 728). Its agents
are that paper’s mind-less agents, at a receiver-side level of abstraction.
It delivers part of what that paper promised. It defines the phenomenon and
gives it a mechanism. It does not deliver the payoff 2004 claimed, a way to
stop looking for the responsible individual. None of the 2004 machinery
(the three criteria, criterion (O), accountability against responsibility)
is used, and responsibility for DM is left to faultless responsibility
(LIT-899), which extends it. The method of levels of abstraction
(LIT-900) is cited twice and not used, so no relation to it is written.
What the reading found on the questions left at seeding:
- Distributed knowledge, not common knowledge. The analogy uses distributed knowledge (Halpern & Moses 1990; Fagin et al. 1995). It is looser than it looks. The paper makes distributed knowledge depend on interaction, which the cited semantics does not. Its examples sum like contributions past a threshold, where distributed knowledge combines different contents logically. Whether the 2016 paper’s common knowledge is a substantive change is for that paper’s reading.
- The network is called an agent, without a test. C = AB is a “supra-agent” by stipulation, and corporations, the State and even moral enablers are “agents” (pp. 729, 740–741). No criterion is applied. The paper does not bear on QUESTION-009 beyond showing how freely the line uses the word.
- Infraethics is less neutral than the abstract says. The body claims only that it is not necessarily good. It concedes that separating it from values “never occurs in reality” (p. 739), and it lists values (privacy, transparency, freedom of expression) among its enablers.
No THEORY is filed from it. The record files none for the comparable conceptual readings in this line.
Access when seeded: Crossref gave the metadata and the reference list, with no abstract. The PhilArchive copy (philpapers.org/archive/FLODMI.pdf, green open access) returned a Cloudflare 403. The University of Hertfordshire record (uhra.herts.ac.uk/id/eprint/20632, to which the older handle 2299/11843 redirects) has no file. The Springer version is not open access, and PubMed (23197312) has no free text. The owner supplied the published PDF on 2026-10-10; it is not held in the repository. published: is the online date Crossref gives.