TIL you can use Boolean operators in Iris.ai if you know where to look.

I've been evaluating Iris.ai for the past several weeks, primarily with an eye towards how its search and filtering capabilities could assist in compliance-driven literature reviews—think SOX control frameworks or HIPAA implementation studies where you need to prove your search methodology is both comprehensive and precise. Like many, I started with the basic keyword searches and the "context filters" they prominently feature. It was useful, but I kept feeling the need for more granular control, especially when trying to exclude specific domains or combine concepts in a strict way.

On a hunch, given my background in parsing audit log query languages, I began experimenting with the search syntax directly in the main search bar. It turns out there is a somewhat hidden, but fully functional, Boolean logic layer underneath the GUI. It's not documented in the primary user guides, but it responds to standard operators. This discovery fundamentally changed the tool's utility for me.

Here is a concrete example of a search string I constructed to find papers on data anonymization in healthcare, while explicitly excluding any studies focused solely on the GDPR (as I was reviewing for a pre-GDPR period):

```
("data anonymization" OR "de-identification") AND ("healthcare" OR "medical records") NOT GDPR
```

You can also use parentheses for grouping, which is critical for building complex queries. Another use case I tested was isolating specific methodologies:

```
("systematic review" OR "meta-analysis") AND ("cloud security") AND ("audit trail" OR "logging")
```

Important nuances I've observed through testing:
* The operators must be in ALL CAPS: `AND`, `OR`, `NOT`.
* Quotation marks for phrases are respected and highly recommended.
* The Boolean logic appears to be processed *after* the initial AI-powered concept matching, meaning it filters the Iris.ai results set. It's a hybrid approach.
* There is a limit to the complexity, and very long strings might time out. I've had the most success with 3-4 grouped concepts.

For anyone using this for compliance or systematic review work, this functionality is a game-changer. It allows you to document your exact search string as part of your methodology, which is a requirement for many audit trails. You can move from a broad, AI-suggested concept map to a replicable, precise query. I am now running parallel searches: one with the full AI context filters for discovery, and one with a locked-down Boolean string for the definitive, auditable search. It does make me wonder what other advanced syntax might be lurking in the platform—perhaps proximity operators or field-specific searches? I'll be continuing my experiments.