In October 2018, U.S. District Judge Christopher Cooper ruled that “FOIA requests are not a game of Battleship. The requester should not have to score a direct hit on the records sought based on the precise phrasing of his request.” In an opinion which cites Summers v. Department of Justice and Bagwell v. Department of Justice, Cooper ordered the Department of Homeland Security to conduct an additional search of agency records in response to the Government Accountability Project's FOIA request. The judge ordered a new meet-and-confer within two weeks of the ruling to “engage in a good faith effort to arrive at a reasonably limited set of additional search terms that rectify the under-inclusivity of the 'ideological tests' and 'cellphone' search terms without being too over-inclusive.”

The opinion forms the third ruling in the District Court for the District of Columbia which upholds the Freedom of Information Act and orders expansion of search terms from FOIA requests issued to government agencies. As a result, we can expect that other agencies will be subject to the same standard in the future. This means government agencies and their partners must be prepared to conduct defensible searches and show that they have made a good faith effort to respond.

As Cooper noted in his opinion, under-inclusive terms may be tantamount to asking the responding party to prove or disprove your case through the search, which is a tough standard for any organization. Both parties should think expansively about what defines the issue at question—in this case, an “ideological test.” Define a universe of terms which are relevant to the query—what you might be looking for may need to broaden beyond the specific language used in the request to could include words that suggest conversations about, for instance, political viewpoints, religious affiliations, worldviews, etc.

There are some relatively basic technologies available to accomplish this query expansion. Parties can use software tools to augment the defined terms—political, religious, etc. Identify the word roots for those words (Islam vs. Islamic) and stem the terms to find every variant (Islamophobe, Islamicization, etc). Next, expand into synonyms (Muslim in addition to Islam, Progressive in addition to Liberal, etc.) to cover potentially relevant words or phrases.

An unsupervised machine learning technique known as “clustering” uses patterns in language to identify related topics/domains. Clustering technology will organize documents into high-level themes, making it easier to isolate irrelevant docs, such as internal HR matters, etc., which are divorced from the cause of action. Clustering can help identify thematically related documents, such as discussions of criteria for entry, and hone in on those patterns. Clustering technology is increasingly available to get a quick sense of the query or relevance.

Applying these technologies demonstrates to the court that you have applied all available resources to the FOIA request, while ensuring that the search terms don't incur the burdensome cost associated with being too over-inclusive, or being asked to prove the requesting party's argument.

Once the clusters and query expansion techniques have been applied, and responsive documents are identified, supervised machine learning tools like predictive coding may be applied. These machine learning tools can be trained to identify other potentially responsive documents and help to further weed out the irrelevant documents. Even if your team has not used this technology before, predictive coding may be worth applying for a highly-defensible search. This will further show that the search team has left no stone unturned, and that human error has been minimized.

When these technologies or the budget requirements aren't available to conduct this more comprehensive search, an alternative (or complementary) good-faith response is to ask the requestor for a more carefully-defined FOIA request, and negotiate a list of reasonable terms. Parties who run a search on a limited and under-inclusive list of terms may expect that a judge will recognize when they are being purposely obtuse. Whatever approach you take, make a visible effort to understand the requesting party and their intent at a deeper level. Ensure the search reflects how real people discuss the issue at hand, to validate that your search and response are conducted in a defensible manner.

Jon Kerry-Tyerman is Vice President of Business Development & Intelligence for Everlaw. Previously, he served as a Senior Director in the Innovation practice at LexisNexis, where he chaired the Digital Culture Task Force. Jon served over eight years as a Professor of Law at the University of San Francisco, where he worked in the Internet and Intellectual Property Justice Clinic.