Noggle, R. (2018; substantive revision 2026). The Ethics of Manipulation. Stanford Encyclopedia of Philosophy. https://plato.stanford.edu/entries/ethics-manipulation/.
Key takeaways
From the reading, NOTE-744.
- Two questions. What manipulation is (identification) and whether and why it is wrong (evaluation). The answer to the first constrains the second.
- Every single-mechanism account has a counterexample.
- Covert influence (Ware, Strudler, Susser et al., LIT-1002) misses peer pressure, nagging and guilt trips, which are open.
- Covert intent misses the teenager whose nagging for a phone is “quite transparent”.
- Non-rational influence catches graphic smoking warnings and dressing up for an interview.
- Trickery misses charm offensives, nagging and emotional blackmail.
- Pressure misses pure misdirection.
- What is left. A disjunction (Rudinow: “deception or pressure or … playing on a supposed weakness”), a common factor (Noggle’s 2025 Mistake Account; Klenk’s carelessness about whether a method reveals reasons; Baron’s vice), or no definition (Ackerman; Cohen’s cluster concept).
- Wrongness. Not wrong by definition or absolutely; the entry leans, with “perhaps”, to prima facie and pro tanto. Harm cannot be the only wrong-maker, since beneficent manipulation seems wrong. The autonomy account faces autonomy-enhancing manipulation and S. Buss’s argument. Renzo’s account, on which the manipulator co-opts the target’s own practical agency, is given most room.
- Intent is open (§4.2). Baron allows intent with recklessness, Manne manipulation without conscious intent. For an algorithm that selects interface variants which “had they been chosen deliberately by a human, we would label as manipulation”, five answers are listed and none endorsed.
- Ordinary, not global. The free-will literature’s manipulation is “radical programming or reprogramming”, and “few people have explored the connections” with ordinary manipulation.
Standing in the record
Filed on 2026-10-12 at the owner’s request with QUESTION-070 and QUESTION-071, and read in NOTE-744. It is held as the record’s map of the analyses of manipulation, to which those questions appeal. Its one conclusion with moderate support, that no single-mechanism analysis fits the clear cases, is filed as THEORY-225, and its objections set the scope of THEORY-235.
For QUESTION-070, its §4.2 asks the question’s first part for a learning system and lists the candidate answers, including the two branches the question names. It does not discuss a network with no designer. For QUESTION-071, it is about ordinary manipulation and says that its relation to the global manipulation of the four-case argument (LIT-1006) has hardly been studied. So it does not by itself supply what that argument assumes. The 2019 paper by Susser, Roessler and Nissenbaum (LIT-1002) takes its classification from the 2018 version and is quoted in this one.
No instruction for machine-learning practice; nothing here belongs in the anthology.