FBI file 4543-4551 is part of a FOIA obtained on-line collection “United States Military Medicine in War on Terror Prisons,” at the University of Minnesota’s Human Rights Library. That document, “Interviewee Interrogation Plan,” contained the history of one Guantanamo Bay (GTMO) inmate, the methodology by which (presumably) his interrogation will be conducted, and further recommendations. Given that in the first three years of operations roughly 24,000 interrogations were conducted at GTMO, and that all organizations involved in the interrogations are firmly embedded in a hierarchical culture reinforced by documentation practices, one has to believe that each of these interrogations involved documented plans. That trove of possibly 200,000 pages of methodological documents would reveal compelling aspects of the collective history of the inmates, the evolution of interrogation practices during the first phase of the “War on Terror,” as well as present an opportunity to use text analytics to chart the development of concepts and the exchange of language between the networked community of prisoners, guards, politicians and others, that was GTMO.
That is, this research would be possible if the documents weren’t redacted. “FBI 4543-4551″ is one example of redaction gone to an extreme. In 9 pages, only the following 29 words remain.
“SAs  and  have established a well-organized and thoughtful blueprint for the interview and interrogation process for this detainee. They requested the assistance of SSAs  and  on the following matters:” (p. 4547).
What remains is something like government self-affirmative poetry — a piece of condensed writing lauding an invisible set of decisions. This condensed, redacted form enables a kind of close reading more expansive documents resist; i.e., interrogatory conversations are considered technical structures like buildings and machinery to be conducted according to a formal mechanical rigor schematized in “blueprints,” much in the manner of how rhetorical figures were deployed in classical speeches.
What was added to the document might be even more compelling. The gutters of the document now contain a nearly unbroken column of codes:
Those codes correspond to FOIA exemptions allowed in 5 U.S.C. 552(b) to federal agencies releasing documents. At each redacted element, the redactor provides a code indicating the reason for keeping that information out of the public domain. See Phil Lapsley‘s “How to Read an FBI File” on http://www.historyofphonephreaking.org/writings/htraff/ for a more complete description of this process, and “XI. Summary of Exemptions” at http://archive.eeoc.gov/foia/hb-11.html for a descriptive list of reasons why a given piece of text was elided. Combining XI with the above list gives:
Relates to internal personnel rules
Intra-agency communication that would not be available to a litigant
Personal information of a private nature
Would invade the privacy of a law enforcement officer
Would reveal the identity of a confidential informant
Would reveal law enforcement techniques the knowledge of which might help people to break the law
Protect the safety of a person [originally, limited to the safety of law enforcers]
This configuration of redaction codes, rather than unique to a given piece of redacted material, are repeated verbatim 32 times in 9 pages of a document conceivably much longer than 4,551 pages. Their repetition provides an interpretable moment, and an opportune moment for a text analytics approach. Compiling code frequency, pattern, and correspondence information could reveal cognitive and work patterns of federal redactors of witness testimony. Along with allowing for a demonstration as to whether the codes are diligently or lazily applied, the specific configuration of codes and the relation of codes to sections might present a constellation better describing the networked community of sites of interrogation. That constellation emerges because of patten-based questions like, does b7C always trigger b7D, or only a percentage of the time, and only for certain types of documents?, does detainee background always trigger codes related to the safety of prisoner and interrogator?, does it trigger concerns about litigation against the agency? Those questions are feasible because of computational approaches to the understanding text system practices, like redaction.