Library News

🌈 The Diversity Audit: Unpicking Policy, Painstaking Process and Helpful HAL 🤖

Charlie Parkinson, English Lecturer and Library Learning Facilitator at Bradford College

Hello again! In my first blog – published in December 2024 – I explored the importance of representation, and why it is desirable for a library’s collection to mirror the demographic of its users. At that time, I had recently begun working on Bradford College Library’s Diversity Audit. The aim of the project is to determine whether the current stock is appropriate for our users in terms of its diversity of representation.

In order to build a collection that not only represents our users but allows them to explore a variety of alternative perspectives, it is necessary to conduct research into the authors and content of the titles in our existing collection. What does this process look like in practice? This blog will examine Bradford College Library’s current collection development policy, before detailing practical considerations of the task, and offering advice on befriending generative artificially intelligent chatbots. 🤖

Unpicking Policy

I’m keen to discover whether our library collection development policy has helped us to grow a diverse reading collection, or if representation is skewed in favour of the historically dominant social group: *take a deep breath*…white, European, economically privileged, able-bodied, cis-gendered, heterosexual, neurotypical men.[1]

According to our current Library Stock Policy, ‘The Library will endeavour to provide resources in sufficient range and quantity […] to support the courses on offer by the College.’ It is the Academic Liaison Librarians (hereafter referred to as Librarians) who act on behalf of the library to provide users with resources and information. It’s the Librarians’ responsibility to ‘make informed decisions about purchases, within budget restrictions, based upon their knowledge of the subject, the needs of individual departments, and the suitability of the resources under consideration.’[2]

Five priorities for the selection of material for inclusion in the library collection are listed in the Library Stock Policy; although one of these is ‘Compliance with other College policies’, there is no explicit reference to equity, diversity or inclusion.[3] For this reason (and because our entire socio-economic system in Britain is underpinned by inequality – see feudalism, social class structure, colonialism, the slave trade, patriarchy etc.), I suspect that we will find a representational imbalance – in favour of that aforementioned historically dominant social group…but I hope I’m wrong.[4]

To prevent (or at least minimise) discrimination and tackle institutional inequality in any organisation, it is necessary to first acknowledge that discrimination has happened throughout human history and is still happening today. A one-off quick-fix does not exist, so on-going training is essential. Expectations need to be clearly communicated to staff and patrons; those who raise concerns must be listened to and supported; policies and practices should be regularly reviewed and updated.[5]

Once we have completed The Diversity Audit of General Fiction, we plan to embark on a larger-scale audit of the entire library collection. Following – or perhaps during – this process, I believe it will be necessary to assess and revise our library collection development policy to ensure that equity, diversity and inclusion are explicitly stated as priorities. We must then carefully consider how to address systemic bias – if (or when) it is revealed through the audit. I predict that this will be the trickiest stage.

You may be wondering what exactly is involved in auditing a library’s collection for diversity; this next section offers an insight into the evolution of the process thus far. 

Painstaking Process

Firstly, our Library Team Leader created an Excel spreadsheet, listing the entire Reading Collection – alphabetised by authors’ surname. There are 939 titles in total. I chose to consider the authors before content and found that there were 671 different individuals to investigate.

I planned to begin by researching each author to determine whether they publicly disclose(d) their ethnicity, nationality, religion, gender, and sexuality, or whether they have revealed any disability, neurodivergence, or adverse socio-economic experience. Unless authors have unambiguously referred to these aspects of their identity in the public domain, or there are verifiable sources (such as historical correspondence) readily available, I must leave boxes unchecked. I believe that, whenever possible, assumptions regarding a person’s identity should not be made, so if they haven’t revealed that information to the world, it’s not getting added.[6]

Each of the categories on the spreadsheet, has its own drop-down list of sub-categories. For example, if I click ‘gender’, I can choose between ‘female’, ‘male’, ‘transgender’, ‘non-binary’, or ‘other’. The categories and sub-categories have been the source of much discussion and debate. They are fluid, meaning that they are under constant review; we are not resistant to changing them, or adding to them, if necessary. Categories – of probably anything, but especially of people – are a lot more complicated than they seem at first glance, so my next blog will be entirely dedicated to categorical conundrums.

Just how was I going to retrieve all of that information about each author?

Here’s a library life hack for struggling English students: on the Good Reads website, you can search a huge range of books to find plot summaries, community reviews, statistics, information about author, genre, editions and recommended reads (titles generated by an algorithm based on your previous ratings, reviews and search history). This website would be my first port of call to check the diversity of the authors in our collection.

Whenever I found a dearth of information about an author on Good Reads, I had to sift through other online sources, including Encyclopaedia Britannica, web-articles, interviews, authors’ websites, biographies, and Wikipedia. If I still came up dry, the boxes remained blank – which soon became a source of personal dissatisfaction. This must be how it feels to be a private detective. Perhaps I needed a pinboard, some mug shots and a bunch of red wool...

One colleague asked me if I planned on calling every author and asking them directly if they would please disclose their sensitive information for the sake of my library research project. Although, I admit, I did entertain the fantasy of phone chats with all my favourite novelists for a good 20 minutes, I decided against this method (for fear of lawsuits, bankruptcy and prison). Margaret Atwood though – if you’re reading this – call me, yeah?

Margaret Atwood - Charlie's favourite author and All Round Top Lass

This is how I started out – using any spare time in my working days to painstakingly scan online text, eking out personal information about those 671 writers. I worked out that I could average 6-8 authors in an hour, depending on the wealth of information available and how many times I had to break off the task to fulfil some other duty – or whether I became distracted by random interesting facts! Did you know that D.H. Lawrence apparently had a penchant for climbing mulberry trees in the nude?[7]

D. H. Lawrence - Famous Author; Secret Tree Climber

The plan was that once I had collected as much information about the authors as I could glean from the internet, I would survey the content of each title. I was a little concerned that my Team Leader might ask me to read all 939 books to find out what they are about. Of course, she didn’t demand I do that, but even the prospect of reading 939 bite-sized plot summaries, distilling the vital information, and inputting all of this data was pretty daunting.

If the plot covered issues specifically pertaining to any of the categories listed on our spreadsheet, I planned to go to that category’s drop-down list and select the most accurate description. If the book did not cover any of our categories, but I thought that it did cover another aspect of diversity, I would make a note of this. If a category was not covered by the book, I would leave its column blank.

This was going to take a while!

Thankfully, after a slow start, I discovered a shortcut to make this mammoth task more manageable.

Helpful HAL?

HAL 9000 - The A.I. Antagonist of 2001: A Space Odyssey 

At the start of November, I had a revelation: one way of potentially speeding up the process significantly – and perhaps even capturing more accurate data – was to enlist the help of Artificial Intelligence. In five seconds, Microsoft Copilot could do what it would take me several hours to accomplish – providing I gave it clear, detailed instructions.

Initially, optimistically, I asked Copilot to complete the spreadsheet for all authors in one go. This didn’t work out well. The chat-bot immediately replied, ‘This will take a bit of time.’ When I pressed it to provide a rough estimate of the duration of the task, its response was ‘This task may take over fifteen hours to complete’. Copilot confirmed that this was a hefty chore – even for an incorporeal cluster of advanced technologies masquerading as a friendly online pal. My overwhelm was justified. Phew.

After leaving Copilot running in the background for a few hours, it soon became apparent that this wasn’t going to be as easy as I’d anticipated. It crashed. Several times. I suspect that this occurred whenever the Wi-Fi dropped out, or when the college online security system kicked in, periodically logging me out of all platforms.

After that, I tried a different strategy; this one was a winner. I decided to set Copilot to work on twenty authors at a time. That way, I could easily monitor its progress and save the information in chunks, whilst still being able to get on with other jobs in the meantime.

I was once told, in an A.I. Academy training session, to always thank the bot for its labour – because ‘You may just find that you get better results when you do.’ Creepy much? This calls to mind folkloric warnings about the necessity of leaving a saucer of milk out for the fae. Magical thinking or condescending coding? [8]

Using A.I., it took me approximately eight hours – spread over three days – to collect all the information on the authors. Copilot churned out 31,143 words in the process, which I saved to a separate Word Document. It was satisfying to issue my decree – including the appropriate polite expressions, of course – then sit back and watch as the ephemeral elves mended all my shoes.

What I have been doing on this project ever since is fact-checking and inputting the data. Being a middle-aged member of Generation Y, I am still distrustful – even suspicious – of A.I. (I’ve watched Alien and 2001: A Space Odyssey; I’ve played Portal. I know the cake is a lie), so I wanted to make sure it had done the job properly. I enter the information into the spreadsheet, but I double-check it regularly. I have learnt, over time, to give the technology more credit. The only occasions I’ve had to make alterations so far have been when Copilot’s failed to retrieve any information at all. When that occurs, I delve a little deeper – looking to books, journals and databases – to fill in as many gaps as possible.

By the third week in January, I had completed the information for just over 600 authors on the list. I’m confident that – with the help of Microsoft’s answer to Johnny Five – it won’t take me long to complete the rest. My next task is to work out how I can use this smart technology to accurately evaluate the diversity of content for each title on the list. Perhaps if I burn some incense, festoon my laptop with flowers, and crack out the tea lights our robot overlords will look upon me favourably. *

* Disclaimer: Dear IT Department, Senior Leadership Team, suggestible students, and any other potentially offended / easily influenced parties, this is a joke. I absolutely do not condone setting fire to things…besides, everyone knows our Masters prefer cake.