A Government Librarian at the 2024 UKSG Conference: Part 1, by Naeem Yar

 

Naeem Yar is a Librarian with the Welsh Government, and serves as GIG’s Events Co-ordinator. This is the first in a three-part series of reflections on the 2024 UKSG Conference in Glasgow, which are written in a personal capacity.

“So, what do you do at the Welsh Government?” That, proceeded by a short pause, was what someone asked me after I received my delegate badge at the first-timers’ reception on Sunday evening. It would be a recurring question over the next few days, and in all fairness, an understandable one. I already knew that the Conference was a place centred on academia, where academic librarians could meet with academic publishers and other resource providers, and the former certainly form a large section of the library workforce. Contrastingly, GKIM (the Government Knowledge and Information Management profession, one of 20 recognised professions in the Civil Service) is a relatively small pond and librarians only form a part of it, alongside colleagues undertaking records management, knowledge management and satisfying freedom of information and data protection rights, among other roles.

 

In addition, I have noticed a paradox in the five years I have been working as a Civil Service librarian: that whilst the work we support is often high profile, our own visibility among the wider profession is somewhat lower – something I tried to address in a small way last year in my role as Events Co-ordinator for CILIP’s Government Information Group by organising a panel webinar titled “What's it Really Like Working in Government Information?” (keep your eyes peeled for a follow up this autumn). I am always happy to do my bit to raise awareness and explained that the Welsh Government (and other Civil Service departments) have librarians too, focused on ensuring that our users can find, access and use the information they need to carry out their roles in an evidence-informed way; in fields from healthcare to heritage, across functions including policymaking, regulation, legislation, carrying out research and internal professional services. such as human resources.

 

We are not academic librarians and our user base is different, being focussed more on practical means of gathering information to inform decision making than aiming to seek information on topics in a comprehensive way. Nevertheless, the importance of finding, getting hold of and using high-quality evidence to our patrons means that we are part of the scholarly information community (to borrow the lovely and inclusive phrase used by UKSG themselves) and my interest in attending the conference was sparked by the relevance of many of the sessions to our practice. I would not be disappointed.

 

Glasgow's SEC Centre, our daytime base for two-and-a-half days (Naeem Yar)

Things kicked off in earnest on Monday morning with a thought-provoking plenary on how best to promote academic integrity in the production and publication of research, featuring the perspectives from those involved in research, publishing and scrutinising outputs. The researcher view was provided by Inke Näthke, Research Integrity lead at the University of Dundee, who noted early on that what constitutes responsible behaviour among researchers is not set down in law in the way that, for example, responsible driving behaviour is: it struck me that this might be resisted by academics as an infringement of their academic freedom and professional autonomy.

 

Also interesting was her discussion of PubPeer, a site designed to facilitate “post-publication peer review”. I had heard of PubPeer whilst working with colleagues on developing a teaching session on advanced internet searching, but I had refrained from suggesting it as a resource to signpost to our users to aid with evaluating articles as it allows anonymous comments. Which, I thought, carried a risk of providing a vector for unfounded and possibly libellous, criticism. I was therefore relieved to hear that the site’s emphasis is to enable discussion of the possible shortcomings of research rather than to make allegations of misconduct. I have also learned that the site moderates comments and the guidance for posters, which moderators take into account, stresses the importance of providing substantive supporting information when making critiques (more information can be found in the site’s FAQs). This gives me some reassurance and provides a spur to give the site a second look as a tool to flag up possible flaws in research.

 

The importance of not assuming that flaws in the research are the result of deliberate manipulation was something that Inke emphasised more widely, drawing on her own experience of investigating possible misconduct to highlight the emotional impact of allegations on those on the receiving end of them, as well as being open to the possibility that inaccuracies are the result of mistakes by the publisher as well as the scholar. This echoed a fascinating conversation I had at the first-timers’ reception with UKSG’s own Bev Acreman, who drew upon her own experience in publishing to highlight that errors were sometimes the result of legitimate mistakes or because research approaches used had been eclipsed by progress elsewhere; that this needed to be distinguished from wilful misconduct and that the withdrawal of publications for honest aberrations should not be considered as a mark of Cain, but as part of the process of scientific advancement. Not accounting for the possibility of honest faults could lead to increasing disincentives for researchers to be open and honest, piling suspicion on top of all too human feelings of embarrassment.

 

Inke, also, mentioned how accused researchers were often upset about their integrity being questioned, even when their actions were open to legitimate critique. This points to the importance of professional ethics in socialising and integrating people into their chosen profession, and a fear of transgressing these ethical codes resulting in a severe blow to the individual’s self-image as well as serious career damage. I can see how such dynamics exist across other professions as well: in the UK, both librarianship and the Civil Service have ethical codes (the CILIP Ethical Framework and the Civil Service Code respectively), and I can understand how being accused of violating those could be a painful experience. Particularly if there is an assumption of this being the result of a deliberate decision in order to advance self-interest.

 

Inke was followed by Daniel Hook of Digital Science, who provided the publisher’s perspective, albeit via video link rather than in person after his flight to Glasgow had to be grounded mid-way due to some serious turbulence. He provided a “big picture” outlook, highlighting the importance of mutual respect for norms and obligations to engender trust, the way our ability to monitor research outputs had been challenged by academic and publishing trends over the last 50 years, and the value of “trust markers” in metadata to serve as tools to help us assess whether outputs are trustworthy. He, also, discussed the potential use of artificial intelligence to analyse networks of researchers in the scholarly community to identify individuals whose publishing behaviours appear to be outliers, such as those who collaborate with far more people or publish far more outputs than the average researcher.

 

Daniel Hook offers a "big picture" in more ways than one. Chair of UKSG and the plenary session Joanna Ball listens intently (Naeem Yar)

Whilst it was acknowledged that such patterns could sometimes be reflective of excellence rather than fraudulence, it seems to me that care should be taken in how usage of such tools should be combined with a human approach to dealing with what is, as has been outlined, a sensitive subject. Though AI is a fast-moving field, a sensible consensus seems to have formed that such tools should be subject to human oversight and scrutiny, and always engaged with critically – something that goes to the heart of librarians’ practice as teachers of information literacy, of the need to unpack the processes leading to a finding as well as examining the wider contexts of the phenomena being studied. I asked a question on the balance between AI use and human oversight in detecting transgressions via the conference app (which unfortunately was not picked up), with particular stress on the fact that many of those who may be caught up would be early career researchers under great pressure to make an impact, working in a system based on incentives that set people up to ”publish or perish”. Just as human factors need to be taken into account when looking to establish systems to strengthen research integrity, so do structures which nudge people towards undesirable behaviours.

 

This point was something discussed by Retraction Watch's Ivan Oransky, who I had also had the fortune of meeting at the first-timers’ reception the night before. He first discussed the progress that publishers had made in increasing retractions and the distance they still had to go, including a need to retract flawed papers more promptly (a problem recognised in the House of Commons Science, Innovation and Technology Committee's 2023 report on Reproducibility and Research Integrity). Then he highlighted structures put in place in China and Saudi Arabia, among other countries, which created pressures that made individuals and institutions resorting to unethical research publication practices more likely, although the Chinese government has taken action to rectify this.

 

After refuelling with one of the excellent grab-and-go lunches provided and exchanging thoughts on the opening plenary with a colleague from Cardiff, who was presenting the following day, I headed upstairs for UKSG’s first workshop on a topic which was front and centre in my thoughts when applying for a place at the Conference. Susan Halfpenny of the University of Aberdeen and Steph Jesper and Siobhan Dunlop of the University of York delivered a two hour interactive session on using critical literacies to teach skills to enable learners to use AI ethically. As I mentioned earlier, I was recently involved in putting together a taught session on advanced internet searching, and as part of this I designed a segment examining the use of generative AI chatbots for information seeking. So, I was keen to learn how academic colleagues undertaking teaching as a major part of their roles were approaching the topic of digital literacy in general and AI in particular.

 

At the heart of the approach taken by the facilitators was the conviction that digital literacy builds on already established literacies. I was pleased to hear this as it tallied with my own experience of developing advanced internet searching training. I was struck by the parallels between the concerns over usage of generative AI and previous anxieties over information seeking using search engines and Wikipedia. It seemed to me that similar approaches could be adopted in developing users’ information skills for using such new technologies, including unpacking how tools produce their results, sources of possible error, their strengths and weaknesses, tactics to use them more effectively and evaluate retrieved information, and highlighting alternatives.

 

These are competencies which can be identified among SCONUL’s seven pillars of information literacy , including scoping, gathering and evaluating information. The seven pillars were originally formulated in 1999 and revised in 2011, and have been used to develop a ”lens” designed to provide a framework specifically for teaching digital literacy, so this approach is well established. This view was also reinforced for me at a recent roundtable event on Information Literacy and AI held by CILIP’s Information Literacy Group, where, in response to the question of whether AI literacy was different from information literacy, there seemed to be a consensus among speakers and attendees that the former could easily be accommodated by the latter. As someone with limited teaching experience, it was reassuring that those with much more practice in the field than I had were of a similar view.

 

I found the first activity in the workshop – coming up with eight points our users would have to consider in their use of AI, with 30 seconds allocated to think of each point – uncomfortable: whilst I understood the intention to be spontaneous about the topic and not over-think things, it was a poor fit for my cognitive style and my brain went on strike after thinking of two points. Nevertheless, the session provided plenty of food for thought, as well as moments where I could draw connections with my own experience.

 

I was lucky to be part of a group where everyone was willing to share their thoughts, leading to a lot of insight and a shared understanding being forged. Note the crumpled remnants of my attempt to "think fast" to the left (Simon Williams)

For example, when discussing chatbot prompts, the facilitators noted that they sometimes returned answers which were tangential to the question asked – something I found when experimenting with the Bing chatbot (now rebranded as the free version of Microsoft Copilot) as a literature searching tool. If I searched for publications on a topic combining several concepts, it would often return documents which were relevant to only one or two of them.

 

Likewise, it was acknowledged that generative AI outputs could be on the bland side, reflecting large language models’ (LLMs) dependence on their training data for learning: what they produce is inevitably derivative, not original. In the same way its answers to informational questions are not going to match those of someone with deep knowledge of a subject. These reflect my own interactions with the technology and that – for the moment at least – it has limits in its use for more sophisticated tasks where a creative and knowledgeable person will be able to produce something better. Frequent changes in the algorithms used by chatbots were also noted, resulting in changes in user experience – something that I have seen in practice, in particular to curb LLMs’ tendency to “hallucinate” or make up factual claims. Group discussions also helped to identify the specific long-standing information literacy skills which were particularly relevant to AI.

 

Following the workshop, I stayed with the theme of AI but changed the context as I attended Jonathan Chipp and James Howells’ session on using technology to assist with data-driven decision making for collections development at the University of Southampton (presentation available via Slideshare). I was particularly interested in this as our library team had undertaken a project to review and weed our holdings in our main Cardiff library during 2023, and I was curious what benefits AI could bring to such a process.

 

Jonathan Chipp outlines the University of Southampton's collection management project (Simon Williams)

I could see some commonalities in terms of the drivers for this initiative and our own weeding work: a decline in print circulation and pressures on the use of space were shared factors, as I am sure they are across many other libraries as well. On the other hand, the need to utilise space for different purposes is less of an issue for us: I know that physical study space has become ever more important in academic libraries over the years, but there is much less demand for use of a library as a workspace for us. Particularly with the Welsh Government keenly advocating for flexible hybrid working, both internally and as a matter of public policy, though patrons do use the library as a workspace.

 

I was particularly curious about Southampton’s identifying of issues with classification among the collection as a problem they believed the project could help with: this was an area where there was room for improvement in our own collection, and more consistency could help to place “like with like” more consistently and therefore enhance findability. One aspect which struck a chord with me was that working with the collection had led to the discovery of historical correspondence about the collection: I manage a project cataloguing the contents of a collection held by Cadw, the Welsh Government’s historic environment service, and have often come across letters, annotations and sketches which contain information about the history of Cadw’s work, and the history of the heritage sector in Wales more widely. As I like to say, sometimes it feels like the collection could do with the care of an archivist, rather than a librarian. That was not the only reassuringly old-fashioned thing about aspects of the project – they used coloured stickers to indicate which items had been reviewed through the process, which is a technique that we used as well – I suppose if it works, it works.

 

Hearing about how computerised analysis was used to propose decisions about what stock would be retained in the library, what would be relegated to a store, and what would be weeded, I could see how such a process could be used to make the process more efficient and consistent. Thinking back to our weeding project however, I found that spending time methodically working across the collection was a great way to familiarise myself with the physical collection (particularly important as I joined the Welsh Government during the pandemic and have carried out my role initially as a homeworker and then a hybrid worker). I think it was beneficial for the whole team to be involved in decision making around the collection, sharpening our skills to this end. I was therefore pleased to hear that the approach taken by the project at Southampton was to treat the initial decisions as suggestions and as the beginning of a conversation rather than a final say.

 

Although as a non-academic library, we do not have a segment of our users, like faculty, that represent an organised and particularly influential body of stakeholders, who we liaise with through permanent structures, if we did adopt such technology it would be important to use it as a tool, rather than completely delegate decision making to it, retaining the freedom to take other factors into account and maintaining review of decisions by staff. That would be consistent with the critical attitude we need to uphold in advising others in the use of AI and with the valuing and cultivation of our own professional skills.

 

The day concluded with a couple of lighting talks back in the main hall. Jenni Adams and Ric Campbell discussed their work at the University of Sheffield advocating for compliance with the FAIR principles (enhancing findability, accessibility, interoperability and reuse) regarding data and software (presentation available via Slideshare). Although their aim of encouraging researchers to act to remove barriers to allowing others to make secondary use of their data is not directly relevant to my role, I am a member of the Welsh Government’s Research Data Access Panel, which forms part of the system of governance for sharing Welsh Government datasets with researchers. There is a focus on ensuring that proposals are appropriate in terms of their purposes and ethical in their approach. As someone involved in managing demand for data available for secondary analysis, it was therefore interesting to hear about efforts to increase the supply of such data. From an ethical perspective it was good to hear that the presenters provided customised advice to researchers for different types of data, including sensitive data. They were willing to challenge preconceptions about what could be shared in a legally compliant and ethical way by stressing the way tools like consent forms could be redesigned to take the FAIR principles into account, if this was done early enough in the research process.

 

Jenni Adams talks about advocating for the FAIR principles to researchers (Simon Williams)

The first day was rounded off by another lightning talk by Simon Bell of Bristol University Press, discussing their response to the UN Sustainable Development Goals Publishers Compact (slides available via Slideshare). The resources discussed in the talk did not necessarily fit the bill of the interests of our user base. Most of the publications I have found, which discuss the SDGs, tend to examine them in the context of the developing world, whilst most (though not all) of the civil servants working on policy that I have encountered, tend to be interested in interventions in the developed world. These would seem to be tested in similar contexts to the UK and therefore more likely to be directly translatable to our context. This talk, however, did spur me to examine BUP’s publications, which include multiple offerings highly relevant to a range of policy areas, including the blog Transforming Society, the journal Futures of Work and over 140 open access books. I will be drawing attention to all of these to colleagues as resources which may be useful for ourselves and our patrons.

 

That was it for the day in terms of work. I skipped the following reception in favour of a cup of tea and a lie down in my hotel room, but I did head over to the UKSG quiz, where I joined a motley but affable crew of academic librarians for an evening of good conversation and refreshments. Unfortunately, the good spirits were not translated into results as we finished in joint last, but this achievement did result in us being awarded actual wooden spoons carved with the UKSG logo. Something we all agreed made for a fine memento of a lovely evening, where everyone treated the quiz with an appropriate level of seriousness – that is, very little.

 

Me and the rest of the team celebrate winning the wooden spoon, with probably much more enthusiasm than the Welsh men and women in this year's Six Nations (Simon Williams)

Thus concluded the first day. If you would like to hear about trends in Open Access publishing, helpful tools for researchers, strategies to teach how best to avoid predatory publishers and more about applying AI to managing collections, watch this space for the second part of my account of the conference.