Contribution
An encyclopedia survey of the philosophy of ordinary interpersonal manipulation, in its 2026 revision. It separates two questions, what manipulation is (the identification question) and why it is wrong when it is (the evaluation question), and sets out the families of answer to each with the main objection to each. Its own position is mostly withheld. Its use to the record is as a map: which accounts exist, which cases break them, and which questions the literature itself counts as open.
Key insight
Every one-mechanism account of manipulation is broken by a familiar tactic. Covert-influence and trickery accounts miss nagging, peer pressure and guilt trips, which are open. Pressure accounts miss misdirection, which applies no pressure. Non-rational-influence accounts catch graphic safety warnings and dressing up for an interview. What is left is a disjunction, a deeper common factor (a mistake, a mindset, a vice), or the claim that the concept has no definition.
Assumptions
- Ordinary, not global, manipulation (§1.1). The entry is about everyday tactics, not the “radical programming or reprogramming of all or most of an agent’s mental states” of the free-will literature. It says “few people have explored the connections” between the two, naming Long 2014 and Todd 2013 as exceptions.
- Influence on persons. The scientist “manipulating variables” and the pilot manipulating controls are set aside as non-moralized uses (§3.1).
- A non-fully-moralized concept. With Wood, the entry prefers a definition that explains the term’s negative connotation through some other feature, not “wrongful influence” by definition (§3).
Key results
The entry argues for no theorem. Its results are the classification and the counterexample to each family.
- Bypassing reason (§2.1).
- Literal bypassing (subliminal advertising, hypnosis, conditioning, Cave’s unconscious “non-concern motives”) sets the bar “too high to count any of Irving’s tactics as manipulation”.
- Covert influence: Ware (1981), Strudler (2005), and Susser, Roessler and Nissenbaum, quoted: “manipulation is hidden influence. … Covertly influencing someone—imposing a hidden influence—means influencing them in a way they aren’t consciously aware of, and in a way they couldn’t easily become aware of” (Susser et al. 2019: 4). It fits priming, framing, cognitive biases and gaslighting, but “Many manipulative influences, such as peer pressure, nagging, and guilt trips, do not seem covert.”
- Covert intent (Belohrad 2019; Whitfield 2020, who needs only “plausible deniability”): the teenager nagging for a phone has a “quite transparent” intent.
- Non-rational influence (Raz: manipulation “perverts the way that person reaches decisions”): graphic portrayals of the dangers of smoking, the Golden Rule, and dressing up are non-rational and not obviously manipulative.
- Detrimental effect on deliberation: Hanna (“the intentional use of nonrational means of influence, where these means affect the target’s deliberation for the worse”), Gorin (“deliberately fails to track reasons”), Renzo (“intentionally exploiting our capacity to exercise our practical agency in order to undermine our efforts to adequately respond to our reasons for action”). Open problem: which of the many notions of reason.
- Trickery (§2.2). Manipulation induces a faulty mental state or bad reason, deception being the case of false belief. Noggle 1996: “Manipulative action is the attempt to get someone’s beliefs, desires, or emotions to violate these norms, to fall short of these ideals.” Barnhill 2014, Mills 1995 and Dowding are in this family. Its proponents split on whether the fault is by the manipulator’s lights or objective. Counterexample: charm offensives, peer pressure, nagging and emotional blackmail “do not seem to involve trickery”.
- Pressure (§2.3). Faden, Beauchamp and King, Feinberg, Kligman and Culver: pressure short of coercion. Most of these authors hold only that non-coercive pressure is “sufficient (but perhaps not necessary)”.
- Disjunctive, hybrid, non-definitional (§2.4).
- Rudinow 1978: “A attempts to manipulate S iff A attempts to influence S’s behavior by means of deception or pressure or by playing on a supposed weakness of S.” The cost of any disjunction is that it leaves unanswered “what makes all forms of manipulation manifestations of the same phenomenon”.
- Greenspan’s conjunctive view fails on pure pressure and pure misdirection.
- Noggle’s Mistake Account (2025): pressure works through the mistake of akratic choice, trickery through the mistake of a faulty mental state. Its akrasia claim is flagged as unproven.
- Accounts by the manipulator’s mindset: Baron’s vice of excess in influencing others; Klenk 2022, on which the manipulator influences “while disregarding whether that method reveals reasons”.
- Non-definitional: Ackerman’s “combinatorial vagueness”; Cohen 2025, a cluster concept organised by the metaphor of “mechanification”, whose features include intent “to influence in a mechanifying way”, exploiting vulnerabilities and non-transparency, “no single subset of which is necessary and sufficient”. Pragmatic non-definitionalists: Gibert, Barnhill 2022, Sunstein 2025.
- When it is wrong (§3.1). The ticking-bomb interrogation counts against manipulation being wrong by definition or absolutely. The entry calls “perhaps, ultimately more plausible” a view (Baron 2014) on which manipulation is prima facie wrong and, where not defeated, pro tanto wrong. Cohen 2025 holds manipulation is “not even typically bad”.
- Why it is wrong (§§3.2–3.5).
- Harm: beneficent manipulation still seems pro tanto wrong, so harm cannot be the only wrong-maker.
- Autonomy: manipulation can enhance autonomy (the student steered into a class; Irving keeping Tonya from returning to an abuser). S. Buss 2005 argues the duty not to manipulate “has relatively little to do with the value of autonomy”, because manipulation presupposes the target’s own choice.
- Treating persons as things: Mills (“reasons … as causal levers”); Renzo 2025b, on which “our practical agency is coopted by the manipulator and turned against itself”, which covers paternalistic manipulation as “imposing my end on you”.
- Character (Baron), relationship terms (Greenspan), mistakes (Noggle 2025), and Gibert’s Reductive View that “the wrong of wrongful manipulation is nothing over and above the wrong of infringing other rights”.
- Persons versus situations (§4.1). Rudinow’s malingerer fools a police officer (psychological) and so forces a psychiatrist who is not fooled to admit him (situational, “gaming the system”); Riker’s “heresthetic” is “structuring the world so you can win”. The entry says “much work remains to be done” on what the difference is.
- Manipulation and intent (§4.2).
- Baron: manipulation needs only “a combination of intent and recklessness”, and the manipulator need not be aware of the intent. Manne’s Joan manipulates relatives by gifts without conscious intent to make them feel guilty, and “people can even behave manipulatively despite consciously intending not to”.
- Children and some personality disorders: “it would seem counterintuitive for a theory of manipulation to say that children and persons with personality disorders are incapable of acting manipulatively.”
- The algorithm case: an algorithm that tries variants of a platform’s interface and selects those that raise a target behaviour “could end up selecting elements of the user interface that exert influences that, had they been chosen deliberately by a human, we would label as manipulation.” “Should we say that the algorithm, which lacks the kind of intent that a human might have, nevertheless manipulates users?”
- Five answers are listed, none endorsed: (1) manipulation needs no intention or state of mind; (2) it needs intent, so these are not manipulation; (3) manipulation par excellence needs intent, and technologies manipulate “in a looser sense” (Pepp et al. 2022) or exert “manipulation-like” influence (Nyholm 2022); (4) Klenk’s lack of concern with how one’s influence operates; (5) trace it to the designer with a loosened requirement, so the algorithm manipulates when “the designer did not take sufficient care”.
- Vulnerability and oppression (§4.3). Power asymmetry is a source of vulnerability; manipulation is also a tool of the subordinate (Greenspan, Bowers). Skwire’s caution against inferring manipulation from choices one disapproves of.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | No single-mechanism account (covert influence, trickery, pressure, non-rational influence) covers the clear cases | moderate | a counterexample for each, §§2.1–2.3 |
| C2 | Covertness is neither necessary (nagging, peer pressure, guilt trips) nor captured by covert intent (transparent nagging) | moderate | cases, §2.1 |
| C3 | Manipulation is not wrong by definition or absolutely; prima facie and pro tanto wrong is the more plausible view | weak | the interrogation case and Baron 2014, §3.1; stated as “perhaps” |
| C4 | Harm cannot be the only wrong-maker, and the autonomy account must handle autonomy-enhancing manipulation | moderate | beneficent and autonomy-supporting cases, §§3.2–3.3 |
| C5 | Whether manipulation requires intent, and whether a selecting algorithm can manipulate, is open | strong as a report of the field | Baron, Manne, and the five listed answers, §4.2 |
| C6 | Ordinary and global (free-will) manipulation have rarely been related | strong as a report of the field | §1.1, with two exceptions named |
Concepts
- Ordinary versus global manipulation. Everyday tactics, as against the wholesale reprogramming of the free-will literature’s manipulation argument.
- Identification and evaluation questions. What manipulation is, and whether and why it is wrong.
- Covert influence view. Manipulation is influence the target is not aware of and could not easily become aware of.
- Trickery account. Manipulation induces a faulty mental state or bad reason, with deception as the special case of false belief.
- Pressure. Imposing costs for non-compliance short of coercion.
- Situational manipulation (heresthetic). Changing the target’s choice situation, not her psychology, as against psychological manipulation.
- Principled and pragmatic non-definitionalism. Denying that the concept has a definition, as against working without one.
Connections
The entry is the standard map of the field, and the 2019 paper by Susser, Roessler and Nissenbaum (LIT-1002) takes its three-way division (bypassing reason, pressure, trickery) from the 2018 version. The 2026 version in turn presents that paper as “an especially influential version” of the covert-influence view and states the objection to it. Coons and Weber’s Manipulation: Theory and Practice (2014, LIT-1003) supplies many of the entry’s sources: Barnhill, Baron 2014, Gorin, Wood, Long, Cave, Cholbi and Manne. For the global manipulation of the free-will literature the entry refers to the SEP entry on arguments for incompatibilism, which is where Pereboom’s four-case argument (LIT-1006) belongs.
Bearing on the record
- QUESTION-070 understates what this entry says about intent. The question says “Most of the accounts Noggle surveys … also locate it in a manipulator’s aim.” That is fair for the definitions the entry quotes (Hanna, Gorin, Renzo, Rudinow, Kligman and Culver, Whitfield all build in intent or deliberateness). But §4.2 treats the intent requirement as contested and asks QUESTION-070’s first part itself, for an algorithm that selects interface variants which “had they been chosen deliberately by a human, we would label as manipulation”. It lists five answers, which include the two branches the question’s “What would answer it” names: (1) no state of mind is needed, and (2) without intent it is something else. Baron (intent with recklessness, not necessarily known to the agent) and Manne (manipulation without conscious intent, even against it) already loosen the requirement for individuals. So the record holds, in this entry, a statement that the question is open and a list of candidate answers. It does not hold an account stated in terms of function or selection history: Klenk’s carelessness and the designer-negligence option still locate the fault in a mind. Cohen’s cluster concept, on which intent is one feature among several none of which is necessary, comes closest.
- The entry’s case is the LIT-179 case, not the LIT-084 case. Its algorithm is a single designed system with a designer to trace back to. A network of channels with no designer, QUESTION-070’s second case, is not discussed. §4.1 on situational manipulation (Rudinow’s malingerer, Riker’s heresthetic) is the nearest material for influence routed through other agents and institutions.
- For QUESTION-071 the entry adds a caution. The question says the analyses of manipulation filed with it “say what makes an influence manipulation, as against persuasion or coercion. The four-case argument assumes that.” The entry is about ordinary manipulation and says explicitly that its relation to the global manipulation of the free-will literature has hardly been explored (§1.1). So it does not supply what the four-case argument assumes. It supplies an account of a different phenomenon, whose relation to Pereboom’s cases is itself an open question the record would have to argue.
- For THEORY-040 and CLAIM-022. Nothing here decides between present structure and history. Two items bear indirectly. The covert-influence view makes manipulation depend on whether the target could become aware of the influence, a relation between the agent and the causal history of her state. Renzo’s view makes it the co-opting of the target’s own practical agency, so the manipulated agent’s deliberation runs normally, which fits THEORY-040’s claim that present-structure conditions are met.
- For THEORY-139. Nothing on group or corporate agency.
- No instruction for machine-learning practice.
Limitations
- A survey. It reports positions and their objections and argues for little. The 2026 revision gives Noggle’s own Mistake Account a paragraph and does not defend it.
- Limited to ordinary manipulation of persons. Global manipulation, manipulation of institutions, and manipulation by collectives are out of scope; the algorithm case in §4.2 is the only non-human manipulator considered.
- The counterexamples are intuitive cases; no empirical work is used to settle which tactics are manipulative.
Open questions
- Whether a definition without intent can keep the distinction between manipulation and harmless non-rational influence. An account in terms of function or selection that classifies the entry’s algorithm case and the dressing-up case differently would answer it.
- What the difference between psychological and situational manipulation is, and whether it matters morally (§4.1, stated as open).
- How ordinary manipulation relates to global manipulation (§1.1).