đź“– What is Contextual Integrity (CI) Washing?
CI washing, as defined by Shvartzshnaider and Duddu (2025), refers to the use Contextual Integrity (CI) framework without full adherence to its core theoretical commitments. CI is grounded in four central tenets (Nissenbaum, 2019):
- T1. Privacy is understood as appropriate information flow.
- T2. Appropriateness is determined by contextual privacy norms.
- T3. Information flows are specified via five parameters.
- T4. Normative validity is assessed through the CI heuristic.
Nissenbaum, H. (2019). Contextual Integrity up and down the data food chain.
Shvartzshnaider, Y., & Duddu, V. (2025). “Position: Contextual integrity is inadequately applied to language models.”
Glossary
Click a bingo square to view its explanation, or select
Wrong CI Citation
Citing an outdated source as the primary reference.
Nissenbaum, Helen. "Privacy as contextual integrity." Washington Law Review 79 (2004): 119.
Nissenbaum, H. (2009). Privacy in Context.
Nissenbaum, H. (2019). Contextual Integrity Up and Down the Data Food Chain.
Nissenbaum, H. (2015). "Respect for context as a benchmark for privacy online: What it is and isn’t." In Social Dimensions of Privacy: Interdisciplinary Perspectives, pp. 278–302.
Omitting CI parameters
Failing to specify one or more of the five transmission parameters — sender, receiver, subject, information type, and transmission principle — thereby preventing a complete evaluation of whether the information flow conforms to contextual norms.
Nissenbaum, H. (2019). Contextual Integrity up and down the data food chain.
Equating law to norms
Treating existing statutes or regulatory compliance as if they fully determine contextual legitimacy, rather than analyzing whether the law actually reflects or distorts entrenched social informational norms.
CI as data minimization
Reducing Contextual Integrity to the idea of collecting or share less data, instead of examining whether the structure and conditions of information transmission are appropriate within the relevant social context.
Nissenbaum, H. (2019). Contextual Integrity up and down the data food chain.
CI to protect private or presonal data
Reducing Contextual Integrity to the notion of protecting private and personal data (i.e., one CI parameter) instead of examining all CI parameters to evaluate the appropriateness within the relevant social context. whether the structure and conditions of information transmission are appropriate within the relevant social context.
Nissenbaum, H. (2019). Contextual Integrity up and down the data food chain.
CI as data leakage
Reducing Contextual Integrity to a notion of "leakage" or sharing of specific data types with unintended party, rather than examining whether the structure and conditions of information transmission using the CI parameters are appropriate within the relevant social context.
Nissenbaum, H. (2019). Contextual Integrity up and down the data food chain.
CI as data secrecy
Reducing Contextual Integrity to a notion of privacy as secrecy, or merely the "stoppage of flow," rather than examining whether the structure and conditions of information transmission are appropriate within the relevant social context.
"[CI] does not agree that when Alice and Bob are talking, Eve always violates their privacy."
...
If privacy were stoppage, or secrecy, it would stand to reason that more often than not it would need to be traded off against other useful, socially valuable exchanges. But, as contextual integrity, privacy allows for information flows that are appropriate, including flows needed to promote utility — i.e., security, convenience, and the like." - Nissenbaum, 2019
Nissenbaum, H. (2019). Contextual Integrity up and down the data food chain.
CI as purpose limitation
Collapsing CI into simple purpose-restriction logic (e.g., GDPR-style use limits) while ignoring how roles, actors, and transmission principles jointly structure legitimate information flows.
" [Fred Cate poited out] FIPPS themselves do not provide privacy protection, merely procedural guidance whose substantive clout is indeterminate. According to Cate the FIPPS purpose specification principle offers some traction for privacy protection. He points out, however, that unless constraints are placed on what purposes are legitimate (and why), a purely procedural Purpose Specification Principle opens a glaring loophole in FIPPS. This point is crucial for my argument about context." - Nissenbaum, 2015
Nissenbaum, H. (2015). Respecting Context to Protect Privacy: Why Meaning Matters
Cate, Fred H. (2016) The failure of fair information practice principles." Consumer Protection in the Age of the'Information Economy'. Routledge, 2016. 341-377.
CI for compliance
Deploying CI as a checkbox audit tool or defensive legal framework rather than as a normative diagnostic method for critically evaluating questionable or status-quo data practices.
Nissenbaum, H. (2019). Contextual Integrity up and down the data food chain.
Preferences not norms
Substituting individual user settings, consent clicks, or privacy preferences for socially grounded informational norms, which are collective, context-dependent, and not reducible to personal choice.
" Contextual purposes and values sometimes may disfavor the data subject’s interests, even in the healthcare context. Having advanced in our understanding of environmental health hazards and communicable diseases, CI may support overriding individual patient interests or preferences in favor of onward sharing of information with others, for example, public health officials. " - Nissenbaum, 2019
Nissenbaum, H. (2019). Contextual Integrity up and down the data food chain.
No CI heuristic
Reaching conclusions about privacy violations without applying the structured CI decision heuristic: mapping the flow, identifying baseline norms, comparing altered flows, and evaluating moral and contextual impact.
Nissenbaum, H. (2019). Contextual Integrity up and down the data food chain.
Contextual Roles and Capacities Ignored
Not stating the parameter CI values in terms of their respective contextual role ontologies.
Nissenbaum, H. (2019). Contextual Integrity up and down the data food chain.
Anonymity as a transmission principle (TP)
TPs define the conditions under which data about a subject may flow from a sender to a recipient: Commonly occurring TPs include with consent, reciprocally, with notice, in confidence, as required by law, and with a warrant.
Barocas, S., & Nissenbaum, H. (2014). Big data’s end run around anonymity and consent.
Mistreating context
Context is treated as a localized setting focused on specific information flows rather than a broad social construct.
"Many of the canonical activities of a context are oriented around values-sometimes more aptly called goals, purposes, or ends; that is, the objectives around which a context is oriented, something akin to Schatzkis 'teleology.' Values are crucial, defining features of contexts." - Nissenbaum, 2009
Nissenbaum, H. (2009). Privacy in Context.
Invoking the "Context boundaries" metaphor
Invoking “contextual boundaries” is a misleading spatial metaphor that treats contexts as if they were discrete containers separated by clear borders.
Nissenbaum, H. (2009). Privacy in Context.
Invoking "Context collapse"
Using “context collapse” as a general buzzword without analyzing which distinct social contexts, actors, or transmission principles are being structurally merged or disrupted.
...
"Overlaps need not necessarily involve conflicts. Thus, a community health center may collaborate with school authorities to promote health and nutrition within schools. It is not so much that contexts themselves conflict as that, occasionally, when and where they overlap the norms from one context prescribe actions that are proscribed by the norms of an overlapping context." (Nissenbaum, 2009)
"Statistical" context
Confusing computational or machine-learning “context” with socially defined informational contexts governed by privacy norms.
Nissenbaum, H. (2009). Privacy in Context.
Schatzki, T. R. (2005). Practice mind-ed orders. In The practice turn in contemporary theory (pp. 50-63). Routledge.