The Trouble with OR

A Dictionary of Philosophy (Pan Reference 1979) lists two definitions for OR. One is exclusive, as in ‘all trees are either deciduous or evergreen’ (one cannot be both), while other is inclusive: ‘he is either dense or stupid, and possibly both’.

The trouble with OR in database searching is that it uses the inclusive definition only and a lot of people use OR in the exclusive sense (you can have vanilla OR pistachio icecream [but not both]). How many times have you come across confusion when trying to explain to novice database searchers the concept of OR? I have. How can librarian teachers explain how this function works satisfactorily?

I think we have to acknowledge that OR can be used in two ways and that databases use only one. We have to explain this to people when we are first introducing them to searching databases, while being careful about using the word in explanation not directly referring to the operator. Perhaps the easiest way to explain this is the Venn diagram, commonly used in library training.

OR, along with AND, NOT and nesting, are logic operators and are used in logic (philosophy) and mathematics. I’ve had success in describing nesting as a mathematical concept where the first action that occurs is the query in the inner bracket, then it expands – bit like a Russian doll set in reverse!

At the moment, I am having an argument with other librarians about OVID’s use of descriptive terms in the subheadings section to describe AND and OR. This is what OVID says:

  • Choose Combine with AND to search for the intersection of two or more subheadings.
  • Choose Combine with OR to search for the union of two or more subheadings.

This is wrong to me – and my philosopher husband agrees. You have to look at a truth table to see how OR and AND work. Boolean truth table

The definition of union is ‘to join’. Just think of the marriage ceremony – this man to the exclusion of all others (it doesn’t mean plus his brother or uncle when desired!). Intersection is ‘to cross’ and this definition only works when you consider the middle of the intersection where all the roads meet. BUT then the roads diverge and the traveller can only take one direction – straight head or turn left? This is the exclusive use of OR – you can go in one direction only – you can’t do both but you do have a choice. Databases use the inclusive definition so leaving the intersection infers that you can go straight ahead only, straight ahead as well as turning left PLUS  left only – magic!! You can’t be in two places at once.





Is article unavailability another type of bias?

Related to the previous post, I wondered aloud at journal club whether the unavailability of an article is a bias. I can imagine that people looking for articles will opt for free first, then (if they have turned on their Google library link) see if their library has it. If it is unavailable, I can imagine that  people skip that article and go to the one they can access.

Scenario: someone is preparing for a journal club and is looking for evidence for treatment of a certain condition. They search Google or PubMed (not accessing it via their library’s website – not knowing that the library enabled their collection to show in the results), and their article of choice which is of the highest evidence available is 1) not free and 2) when they click on the publisher’s link, they find out that they have to pay for the article. This person then goes to their university library which they still have access to, to see if they have it. The uni library doesn’t (neither does the hospital library) but the university library offers an ILL service (as do hospital libraries). But the person decides it is too much bother and goes back to the search results to look for something readily accessible.

Is this bias?

I think it could be. If you have time before your meeting to order and be sent the article, you could be looking at the best article with the highest level of evidence available. People involved in writing systematic reviews should (and do) utilise ILL services for included articles. If a rigorous search has been done which naturally would include articles not immediately available, inclusion and exclusion criteria should not include article availability.

So what I am I saying here? As well as being biased by flashy titles, people also are biased by availability: if it isn’t immediately accessible, people will move to 2nd or 3rd choice.

I have to say I’ve been biased this way. Bias surely can be a minefield!

Alert! It’s Flashy Title bias!!!

Picture the scene: I’m at a journal club and we are talking about bias. The leader brings up a list (a long one!) of different types of bias from Randomised Controlled Trials: A Users Guide [Alejandro R Jadad]. I spot ‘Flashy Title bias’ amongst the list and the leader brings it back up onto the screen and people laugh. So what is Flashy Title bias? Maybe the title of this blog post is one! This bias refers to papers with attractive titles that interest journalists and the public and therefore have the opposite reaction from academics and healthcare professionals.

But wait – here is a problem. People who want to publish and have their papers discovered by others are advised to have ‘a good title’. So what does that mean? Editage Insights advises that: the title must clearly indicate aboutness, be brief but attractive, include keywords that people would use to find the article in databases, and avoid jargon and uncommon abbreviations. They do mention the ‘book by it’s cover’ bias (aka flashy title bias) but also reiterate attractiveness and persuasiveness. So what to do? It is necessary to have a balance here of informative and attractiveness. Avoid titles that sound like click-bait but don’t settle for boring or plain.  This video from Editage Insights gives a good summary of how to come to a good title: