“You don’t bring a knife to a gun fight” is perceived wisdom reportedly gathered from several different movies, the first of which might have been the 1987 film The Untouchables. I remembered it from Butch Cassidy & the Sundance Kid , but when I checked IMDb.com, I realized I’d been wrong. The phrase there was, “There are no rules in a knife fight.” Bringing the appropriate weapon and knowing the rules are important to online research projects; they’re not restricted to film violence.

Drilled into me, back when I was learning about online database searching, was choosing the source most likely to contain the answer. Don’t search for ham sandwiches in a welding database. Don’t expect to find a preprint in a database that only contains peer-reviewed articles. Don’t get angry when your search for a privately owned company comes up empty in a directory of publicly traded companies. When you’re warned upfront about the scope of a source, accept those limitations, and if it doesn’t fit your needs, go elsewhere.

Blaming the user interface when the source lacks the content you’re searching for is fruitless. I know someone who complained about the UI for her failure to find a birth certificate at Ancestry.com. She was particularly incensed that, upon entering the site, she was prompted to create a family tree. She just wanted a birth certificate. But that’s not the main purpose of Ancestry.com. The family tree creation is integral to its mission.

Other things can go wrong, even when you understand the source. I have a copy of my father’s birth certificate. It omits his first name and spells his middle name wrong. Should I blame a database if it fails to find the information that conforms with what I know to be true but isn’t what is reflected in the public record? No.

Rules, in online searching, frequently refer to Boolean logic. A search for two words, assuming a phase search, will be unsuccessful if a programmer has decided to assume an OR relationship between the two. The phrase Butch Cassidy retrieves radically different results from Butch OR Cassidy.

AI technologies, however, are subverting the rules of Boolean logic. With machine learning (ML), systems can correct spelling (Cassaday is automatically corrected to Cassidy) whether you want it to or not. Maybe you really did mean to search for Batch Cassaday. Deep learning can detect the probabilities that will lead to successful search results. This works very well for shopping but not so well for serious research. Information professionals need to be alert to bias. Training sets for AI, ML, and deep learning, if not carefully constructed and maintained, can introduce bias about gender, age, ethnicity, and even intent. Looking at alternative data? Learn how it is derived and what the as sumptions are that drive it. Use, teach, and demonstrate critical thinking skills.

You may not think that online searching resembles a knife fight, but as we enter 2021, do choose your weapons carefully and understand the rules of the source and service you’re using.