Why the Same Exemption Gets Applied Two Different Ways

Because the reasoning is rarely written down. Two peoplereading the same exemption reach different conclusions, and the same person candecide differently six months apart, because nothing carries the earlierjudgment forward. Inconsistency is a record-keeping failure before it is ajudgment failure.

That distinction decides what you do about it. Ifinconsistency is a judgment problem, the answer is training and better people.If it is a record-keeping problem, the answer is capturing reasoning at themoment it is produced, which is a much more tractable thing to fix.

It is rarely dramatic. Nobody releases something that shouldobviously have been withheld.

It looks like one caseworker redacting the mobile numbers ofagency staff and another leaving them, because one treated them as personal andthe other treated them as work contact details. It looks like a paragraph ofinternal discussion withheld as deliberative in March and released inSeptember, because the second reviewer read the same paragraph as factualbackground. It looks like the same standard contract clause redacted in onerelease and visible in another.

Each individual call is defensible. The problem is that twoof them exist.

Two releases of the same material, redacted differently, isthe strongest evidence a requester can have that a withholding was notnecessary. It converts a legal argument about the exemption into a factualargument about the agency's own behavior, and the second is much harder toanswer.

It also compounds. Once a requester has two versions, thenatural next step is to ask for the reasoning behind both, and reconstructingthat months later is exactly the work nobody has time for.

Three structural causes, none of them about competence.

The reasoning is not written down. A caseworkerapplies an exemption, cites it in the log as a statutory reference, and moveson. The reference records what was applied. It does not record why this materialmet that test, which is the part the next person needs.

Nothing surfaces the earlier decision. Even wherereasoning exists, it lives in a closed file that nobody would think to open.The second caseworker is not ignoring precedent; they have no way of knowing itexists.

The volume defeats memory. An experienced custodiancarries a great deal of institutional judgment about how this agency handlesclose calls. That knowledge is real and valuable and it lives in one head. Itdoes not survive leave, turnover, or the twentieth document of the afternoon.

The most consistentrecords offices are rarely the ones with the most experienced staff. They arethe ones where a decision made once is visible the next time the same questionarrives.

Not a memo. Three or four lines, written while the decisionis being made.

The specific material, the exemption cited, the reason thismaterial meets that test in one sentence, and any prior decision it follows.That last element is what turns a log into something cumulative. A note reading"personal mobile, withheld under the privacy provision, consistent withthe July release on the same contract" does more work than a paragraphwritten from scratch.

The test is simple. Could a colleague who was not involvedapply the same reasoning to a similar document next month? If not, the recordis a citation rather than a decision.

Decide the recurring categories once. Most agenciesface the same close calls repeatedly: staff contact details, third-party namesin complaints, internal drafts, pricing in contracts. Settling those asstanding positions removes them from the per-request judgment queue entirely.

Make the standing positions visible where the workhappens, not in a policy document nobody opens mid-request.

Write reasoning at the point of decision. Anythingcaptured later is reconstruction, and reconstruction is both slower and lessaccurate.

Review the close calls, not everything. A shortperiodic look at the decisions marked uncertain does more for consistency thanreviewing the routine ones.

This is the problem Govflowas built to attack directly. In its assistive mode, each proposed redactioncarries the reasoning and the exemption it rests on, and a new proposal ischecked against how the same kind of material was treated on earlier requests.Where the new call diverges from the old one, the reviewer is told before theyapprove it rather than after a requester notices.

That check is the part a manual process cannot reproduce atany level of diligence, because nobody can hold years of prior decisions inmind while working through the twentieth document of the afternoon. Rules anagency settles on can be promoted into standing positions, and promoting one isan approval step in its own right, so the engine's own reach stays under humancontrol. A caseworker reviews and approves every proposal.

There is a second mode that stores no reasoning at all andredacts on exact matches, which some agencies prefer precisely because itcreates nothing for anyone to interrogate later. It is a defensible choice. Itsimply does not accumulate, so consistency stays a matter of memory.

What it does not do is decide whether an exemption applies.That is a legal determination and it belongs to the agency. What changes isthat the determination gets made once, deliberately, and is available the nexttime rather than being reinvented.

Consistency is not the same as correctness. An agency canapply a wrong position with perfect consistency, and doing so simply produces alarger and more uniform problem.

So the standing positions need reviewing, particularly afteran appeal goes against you or the case law in your jurisdiction moves. The goalis not to freeze judgment. It is to stop re-litigating the settled questions sothe genuine ones get the attention.

‍