Author: Jonathan Källbäcker

Going to conferences are always interesting experiences, and as with any endeavour I think it’s important to ask oneself what the point of doing it is. Attending HCII in Gothenburg [1] gave me the possibility to meet up with friends and family, attending ARPPID in London [2] gave me the possibility to acquire some top-notch cookies, and there are not really that many culinary experiences that reaches the hights of fish and chips or halv special med pucko.

Halv special med Pucko – the point of going to a conference in Gothenburg?

Despite the personal advantages when travelling for work, the point is still to work, to present ideas, to gather ideas, but the challenge remains to put these ideas into practice. At HCII, I presented our paper examining stakeholders’ views of potential future technologies at airports [3], while at ARPPID I presented our paper on the collaboration throughout our airport project TARA [4] [5]. During HCII, I also gathered notes and talked to people about improving my own research, new application possibilities at airports, AI in large organizations, AI in teaching, AI in design practice – many different perspectives that can be of use for our research.

And then what? An attempt to answer this question came in the form of ARPPID, where the focus was on papers about closing the gap between interaction design research and professional practice. One challenge I take with me is that it can be multiple and different steps, incentives, and definitions between researchers at universities, researchers in industry, designers, and people who are not designers, making discussions between the fields and roles difficult.

There were also articles and presentations focusing on the misunderstanding of HCI methods, which is somewhat ironic since a core idea of HCI that I’ve taken with me is that systems should be easy to use, and therefore a big part of the responsibility of how the technology is used should be on the designer. This idea should naturally translate to usability and HCI concepts as well, which of course is easier said than done. This taps into the central issue of having common language between academia and industry, which our paper largely is about. Maybe it’s not that important to be academically correct all the time, and letting go of the prestige around that our well-defined and (hopefully) well-cited concepts are being bastardized could be more pragmatic in settings when collaboration with people who are not in research is being established.

What does it mean?!

A more long-term perspective on the question “and then what?” was also discussed in ARPPID, mostly during a panel about the future of HCI in a world of (more) AI. A very interesting part of this discussion were the points that there are a lot of negative consequences that HCI research and practice have led to (one thought – the book The Anxious Generation [6]). P(doom) numbers [7] were discussed, some saying 100, some saying 0 – as well as solutions how to mitigate the challenges that potential existentially threatening technologies, like AI, can pose. However, I do believe that despite the best of intentions there is always a risk that the methods, technologies, and ideas will be bastardized or fill a purpose that was never the intent of the developer.

What I take with me from these conferences are therefore three questions: Who are the users (of the methods or technology etc.)? What is a metric of success? What are the trade-offs?

These questions will have different answers for different people, and the focus may, depending on the person, vary between better collaborations, better research, or managing of existentially catastrophic outcomes. This makes us come back to the original question: what is the point – of going to conferences, gathering ideas, developing methods, or creating technologies? The choice is (still, maybe, kind of, not really, never was) up to us (designers, researchers, practitioners, and other stakeholders and concerned citizens of the world).

References:

[1]: HCI International 2025

[2]: Academic Research and Professional Practice in Interaction Design 2025 – ARPPID 2025

[3]: Exploring the Future of AI, Autonomous Vehicles, and Emerging Technologies in Airport Operations: Stakeholder Perspectives | SpringerLink

[4]: Coming soon.

[5]: https://hto.blog.uu.se/current-research-projects/

[6]: The Anxious Generation – Free the Anxious Generation

[7]: P(doom) – Wikipedia

ChatBPT seems to be the answer to all of our questions today. Or so it seems, given all the ways people find to use the tool for advice. The last couple of months I’ve heard students talking about writing entire assignments, documentarians writing scripts, and people asking it for dating advice, which for me feels downright borderline dystopic.

Inevitably, this technology is used in academia as well, a topic which was discussed in the course Scientific Writing, examined by Dolly Kothawala, this past December. The course was a faculty wide PhD course, meaning that there were primarily students from STEM subjects, a setting that usually makes me and my colleague Andreas Bergqvist with our dubious qualitative methods approach to scientific research feel a bit out of place. Towards the end of the course, we had a seminar where Sarah Shakil asked a couple of questions about the usage and perception of different AI-based tools in our work and writing. Here, I will admit my prejudice towards my dear co-PhD-students. I thought that the optimism would be overflowing, and that people would break out in song praising the almighty Chat-Mini-Me. The following slide represents the actual direction of the discussion:

I came out of the class quite optimistic and inspired. One reason being that the discussion we had during the class was what I envisioned academia to be, a high tempo discussion that moved between present day realism and future dystopia, tossing and turning the different perspectives back and forth. The fact that two of our research groups’ projects are about how these new technologies affect work environment and work engagement was of course a contributing factor too (Current Research Projects – Human Technology and Organisations Research Group).

The main reason for me leaving the class optimistic was that despite people being somewhat towards the negative side, most people did admit using it regularly, yours truly included. “Hypocrites!” some might say, but to me it showed that my dear co-PhD-students are curious but skeptical, understanding the practical usefulness in the technology, but also the risks and downsides, and having people with this mindset in academia gives me hope for the future. All and all, I do share this curious skepticism. On the one hand, it is really convenient to have a grammar checker, a translator, an email diplomatizer. On the other, using texts that are completely Chat-Hihihi generated is plagiarism, and I’m getting somewhat paranoid reading anything nowadays, always suspecting an AI lurking in the shadows.

One of the final questions during the seminar was: “What role do you think AI should play in the future of scientific writing?”. While many attendees answered things like “spelling” and “rephrasing”, the most popular answer was “none”. I would encourage something in between, a certain restrained (some might say lagom) and reflective usage of these tools, with a consideration what is lost, because yes, it is convenient and perhaps more productive, and there are many other tools that makes the writing process easier, but especially for me as a PhD student, I still think the student part of it includes practicing to write properly, and the learning process is fundamental in order to form novel ideas and communicate them gracefully. At the end of the day, I just wish for once that Chat-Arrow-In-The-Knee would answer “I’m sorry, Jonathan. I’m afraid I can’t do that. You clearly need to practice in order to improve your abysmal sentence structure”.

_{Bild genererad via DALL-E.}

I projektet jag är del av har vi genomfört runt 30 intervjuer med olika intressenter inom Sveriges flygplatsvärld för att skapa förståelse kring införandet av teknologi på flygplatser och hur teknologin påverkar markpersonalens arbetsmiljö. För att transkribera intervjuerna har vi använt Open AIs Whisper, som funkar på det viset att man lägger in sin ljudfil i ett program och så spottar programmet ut en textfil med transkriptionen.

Generellt fungerade systemet väldigt bra och minskade tiden som krävdes för att transkribera, men viss handpåläggning krävdes definitivt, eftersom det fanns ett antal återkommande problem som är värda att uppmärksamma. Med det sagt ska det sägas att det finns olika versioner som garanterat påverkar outputen, samt att systemet säkert är bättre på engelska än svenska, men här måste jag erkänna att det inte var jag utan min eminenta kollega Andreas Bergqvist som installerade programmet och bland annat säkerställde att det gick att köra programmet lokalt. Följande recension är skriven av en hobbylingvist och hopplös programmeringsanalfabet som inte har full koll på de tekniska specifikationerna utan bara vill ha ett användbart verktyg.

Först och främst kan det av olika skäl bli väldigt knasiga ord som Whisper föreslår. Ofta var det på grund av att ljudkvalitén i sig var dålig, men även när det kommer till väldigt specifika ord, specifikt till flygbranschen exempelvis, kunde programmet få problem. Amsterdams flygplats Schiphol blev vid flera tillfällen fel då Whisper föreslog att det istället skulle vara ”Skimpol” eller ”Skippholm”, medan ”incheckningsdiskar” kunde bli ”inskräckningsdiskar” och i de flesta fallen lades ett ”r” till i ord som ”banrapporteringssystem” till att bli ”barnrapporteringssystem”.

Ytterligare exempel på sådana enkla felidentifieringar kom ofta i talesätt. ”Goddag yxskaft” blev ”goda ytskaft”, ”Dackefejden” blev ”Daccafaden” och den bästa lilla detaljen av alla – när en respondent sa ”krisen är uppfinningarnas moder” identifierades ”krisen” som ”grisen” och därmed tillkom bilden ovan och min ambition att etablera ”grisen är uppfinningarnas moder” som talesätt. Det förekom även fel som att programmet kunde identifiera helt fel pronomen, exempelvis när respondenten ytterst tydligt sa ”vi” kunde det tolkas som ”jag”. Ordklassen är uppenbarligen korrekt identifierad, men ljudet är uppenbarligen helt annorlunda. Enstaka gånger förekom även fel som att ”vi har endast X” identifierades som ”vi har inte X”, vilket såklart blir problematiskt.

Därutöver tillkom problem som jag antar är inbyggda med en viss medvetenhet. Whisper har till synes ett arbetsminne-liknande sätt att arbeta. Exempelvis, när en respondent på kort tid sa att hen hade haft tre olika ”tjänster” tolkades det tre gånger om som ”chanser”. Första ”tjänst” kan absolut tolkas som ”chans”, men de övriga två lät desto mindre som ”chans”. Så i dess ”arbetsminne” verkar programmet ha identifierat ett ljud som en viss textsträng och identifierade följande liknande ljud som samma textsträng, i ett slags ”hellre konsekvent oavsett rätt än inkonsekvent”- resonemang.

Det som däremot fascinerade mig, utöver den generella precisionen, var de fantastiska sammansatta ord Whisper lyckades få till. ”Arbetsmiljökonsekvensanalys”, ”standardiseringskommissioner”, ”toalettservicechecklista”, och även svengelska sammansatta ord som ”safetychef” var några höjdpunkter inom kategorin sammansatta ord.

Det största problemet med programmet är att vissa osäkerheter och tillägg inte tas med. Det var oftast bara det som identifieras som centralbudskapet som inkluderades. Små uppbackningsord som ”mm”, stakningar som ”eh”, och utfyllnadsord som ”liksom” var ytterst sällan med. Mer påtagligt inkluderades inte det understreckade i följande utdrag: En person pratade om en ”Utbildning som är så jävla lång och så jävla tråkig”, en annan sa att säkerhet ”Det är ju vår högsta prioritet. Det är alltid det som är först.” och en annan menade på att ”Yttersta ansvaret har ju min chef för arbetsmiljö.” Det behöver inte vara fundamentalt väsensskilt från vad respondenten sa, men nyanser och detaljer försvinner.

Så varför spelar alla dessa exemplen som jag har beskrivit roll? Här finns det saker att säga om replikerbarhet och vetenskaplig noggrannhet, men framför allt kan det leda till problem om användare förlitar sig på mycket på program som detta. Bara för att det kan visa upp fantastiska resultat på det stora hela så innebär det inte att systemet till minsta detalj har kapaciteten att producera helt korrekta resultat. Överlag är Whisper ett intressant och hjälpsamt verktyg för att öka hastigheten i transkriptionsmomentet, men än så länge kvarstår behovet för mänsklig handpåläggning, för det finns alltid risk att det är en liten gris som gömmer sig någonstans i myllret av tvärsäkra formuleringar.

It’s been a mere three weeks since I started my PhD position in Uppsala and I’m in Swansea, Wales. The occasion is the conference ECCE (short for European Conference on Cognitive Ergonomics). Oscar Bjurling at RISE (https://www.ri.se/en/person/oscar-bjurling) and I got a paper accepted based on a project we did last year, when I was in the Cognitive Science masters program at Linköping University. “Human-Swarm Interaction in Semi-voluntary Search and Rescue Operations: Opportunities and Challenges” is what we’ve named our paper, and it’s a workshop-based study where we had discussions with experts about potential consequences of drone swarm implementation on search and rescue operations.

Having a paper accepted is all well and good, but it should also be presented. Being that this will be my first conference, I don’t really have a clue about the amount of people who will attend each presentation. I feel like it could either be a full stacked audience and bouquets of roses being handed out to every speaker, or just the one half-sleeping audience member glaring disapprovingly at every one of my attempts at arguing for seeing drone swarms as valuable search and rescue team members. With us being 11^th in a line of 15 15-minute presentations the opening day, there is a definite risk that the eventual flowers will be saved for the keynote speakers.

Nevertheless, a presentation is due, and I think that we as researchers have a responsibility to make sure that the ones who do show up to see our presentation feels like it was worth it. Because if there’s one thing I’ve learned during my brief time as a university-employee is that there’s always something else you could be doing. There will definitely be people there who are stressed about grading papers, writing ethics applications, or other potentially more important stuff than watching our presentation. Now I don’t plan to completely take after the late Hans Rosling and pick up the noble art of sword swallowing for this presentation, partly because of time issues, but also because I couldn’t see the “It [The Sword] is for scientific purposes”-argument going all too well at the security check-in at the airport. However my ambition is to convince at least somebody in the audience that looking into the potential of drone swarms might be a good idea.

Similar thoughts of presentation responsibility struck me when I, in the role of teacher assistant, presented a couple of ethical issues at a seminar last week. Not only could the students probably learn more about the Trolley Problem on Youtube than from me, but I’m actually standing there claiming to know about this subject to the degree that I could be teaching it to university students.

So when preparing for this presentation, I’m being meticulous about representing the thoughts of the experts we talked to correctly, so that I can confidently argue for our analyses and conclusions, while at the same time taking the responsibility of putting on a show seriously. Because if I don’t bother, why should the audience?