To register to the seminar and obtain the link to the call, please fill in this formby Monday 18 March 2024.
20 March 2024– 12pm GMT
Remote – Via Microsoft Teams
Alberto Acerbi (University of Trento), Large language models show human-like content biases in transmission chain experiments
Abstract
As the use of Large Language Models (LLMs) grows, it is important to examine if they exhibit biases in their output. Research in Cultural Evolution, using transmission chain experiments, demonstrates that humans have biases to attend to, remember, and transmit some types of content over others. In five pre-registered experiments with the same methodology, we find that the LLM ChatGPT-3 replicates human results, showing biases for content that is gender-stereotype consistent (Exp 1), negative (Exp 2), social (Exp 3), threat-related (Exp 4), and biologically counterintuitive (Exp 5), over other content. The presence of these biases in LLM output suggests that such content is widespread in its training data, and could have consequential downstream effects, by magnifying pre-existing human tendencies for cognitively appealing, and not necessarily informative, or valuable, content.
Bio
I am a researcher in the field of cultural evolution. My work is at the interface of psychology, anthropology, and sociology. I am interested in particular to contemporary cultural phenomena, and I use a naturalistic, quantitative, and evolutionary approach with different methodologies, especially individual-based models and quantitative analysis of large-scale cultural data. Currently, I focus on using a cultural evolutionary framework to study the effects of digital technologies, and I wrote a book for Oxford University Press: Cultural evolution in the digital age. I am an Assistant Professor in the Department of Sociology and Social Research at the University of Trento, and member of the C2S2 – Centre for Computational Social Science and Human Dynamics.
To register to the seminar and obtain the link to the call, please fill in this form by Monday 29 January 2024.
31 January 2024– 5pm GMT
In person – King’s College London, Bush House Lecture Theatre 3 BH(NE)0.01
Remote – Microsoft Teams
Daniele Quercia (Nokia Bell Labs / King’s College London, United Kingdom), Empowering Cities: Health, Culture, Knowledge, and Resiliency
Abstract
The future of the city is, first and foremost, about people, and those people are increasingly producing a variety of digital breadcrumbs. We will see how a creative use of four datasets can tackle hitherto unanswered research questions. We will see how to:
Track people’s well-being at scale from aggregate records of food purchases https://goodcitylife.org/food/project.php
Quantify the cultural capital of neighbourhoods from geo-referenced pictures https://goodcitylife.org/cultural-analytics/project.php
Predict the innovation success of cities from online records of “who works where” http://goodcitylife.org/cities4innovation/
Profile the psychological resiliency of US regions to COVID-19 from tweets http://social-dynamics.net/EpidemicPsychology/
Bio
Daniele Quercia is Director of Responsible AI at Nokia Bell Labs Cambridge (UK) and Professor of Urban Informatics at the Center for Urban Science and Progress (CUSP) at King’s College London. He has been named one of Fortune magazine’s 2014 Data All-Stars, and spoke about “happy maps” at TED. He was Research Scientist at Yahoo Labs, a Horizon senior researcher at the University of Cambridge, and Postdoctoral Associate at the department of Urban Studies and Planning at MIT. He received his PhD from UC London.
To register to the seminar and obtain the link to the call, please fill in this form by Friday 9 February 2024.
13 February 2024– 3pm GMT
Remote – Microsoft Teams
Mila Oiva (Tallinn University, Estonia), Studying Temporal Changes in Long-Term Audiovisual Data
Abstract
Digitized audiovisual heritage data is seldomly studied in long temporal continuum even though a long temporal perspective could help to explain both specificities of our time, and to reveal longer term continuities as well as short-term caprices. Drawing examples from my studies on Soviet and Estonian newsreels, in this presentation I will talk about the different ways my collaborators and I have been studying temporal changes in audiovisual data computationally.
Bio
Dr. Mila Oiva is a Senior Research Fellow at the ERA Chair for Cultural Data Analytics project at the CUDAN Open Lab at Tallinn University. She holds a PhD from Cultural History with a specialization on the history of Russia and Poland and digital research methods. Currently she runs a project dedicated to studying Soviet and Estonian newsreels in 1922-1997 computationally. Earlier she has been studying the 19th century global news flows, circulation of fake historical narratives in the Russian language internet forum discussions, and the Cold War era transnational information circulation.
The paper is open access and can be found here and the abstract is copied below.
Against a backdrop of widespread interest in how publics can participate in the design of AI, I argue for a research agenda focused on AI incidents – examples of AI going wrong and sparking controversy – and how they are constructed in online environments. I take up the example of an AI incident from September 2020, when a Twitter user created a ‘horrible experiment’ to demonstrate the racist bias of Twitter’s algorithm for cropping images. This resulted in Twitter not only abandoning its use of that algorithm, but also disavowing its decision to use any algorithm for the task. I argue that AI incidents like this are a significant means for participating in AI systems that require further research. That research agenda, I argue, should focus on how incidents are constructed through networked online behaviours that I refer to as ‘networked trouble’, where formats for participation enable individuals and algorithms to interact in ways that others – including technology companies – come to know and come to care about. At stake, I argue, is an important mechanism for participating in the design and deployment of AI.
Greetings readers. On behalf of the ERC SAMCOM project (grant no. 947867), I’d like to take the opportunity to share some of the work we have been doing over the past year. We are a team of anthropologists in the Department of Digital Humanities at KCL, who are currently investigating the moral complexities of digital monitoring across four ethnographic contexts in Europe.
We recently published an entry on ‘Surveillance’ for the Open Encyclopedia of Anthropology. The entry is intended as an assemblage of the anthropological scholarship on surveillance, as well as a reflection on what an ‘anthropology of surveillance’ may consist of. It is open access and we hope it may provide a useful starting point for students and scholars alike .
Mikkel Kenni Bruun and Claire Dungey, together with Rose Powell at Newcastle University, have recently recorded a conversation on ‘Surveillance and Care’ for our podcast series on Spotify. You can listen to the podcast, as well as previous episodes, here:
My students, like many others, have noticed the power of artificial intelligence. Let me put it like this: student essays read well these days! But as a researcher of education technologies and their business models, and as a teacher, I can’t help but ask: what does the usage of automated reasoning do to the student experience, and to the human experience?
I view my role as an educator as being one of supporting students in becoming the person they want to become, before they necessarily know what that looks like. This is the emerging process of developing free and autonomous individuals through learning. At its heart sits a pedagogical paradox that has shaped the institution of education: autonomy for the individual can be reached through socialisation into larger institutional norms and knowledges. In higher learning, we do this by introducing students to bodies of knowledge that we read, write about, and debate. The cherry on the cake is all the interesting things students do outside of the classroom in sports clubs, student societies, employment, their living arrangements, and much more. You know, living!
At the risk of sounding cliché: it really is about the journey of becoming. Becoming a person better able to express and participate in a wider range of the human experience. Let me illustrate this with an example from a popular YouTube channel hosted by vocal coach Cheryl Porter, who films coaching sessions with her students. The sessions are performative and their production is steeped in the commercial logics of the attention economy. Some videos, for example, mirror the well-known format of James Corden’s Carpool Karaoke. But through all the performance and framing, you can also see a process of becoming. In the clip, 11-year-old Isabella is learning to deepen her vocal cord control in order to master Adele’s ballad Easy on me. Adele reportedly wrote the song as a way to process her divorce. In the clip, you see how they work on vibrato consistency and frequency, vocal breaks, and more.
What you see in the video is repetition on the part of the student, and guidance on part of the teacher. Through this learning, social bonds are created, experiences are formed, and both are in a process of becoming.
We can imagine a world where students of music can use software to digitally achieve similar outputs through digital techniques such as pitch corrections. Similar output, however, does not equal similar learning, and thus opportunities for becoming. Put another way, the session between Isabella and Cheryl matters as more than just an output. It matters that the student learns to sing by herself, with her own lungs and vocal cords because the process opens up opportunities for personal transformation. The student’s embodied knowledge of the art of composition will deepen, as will her personal relationship to the emotional states and enduring themes that the song conveys: learning how to do things by yourself shapes who you are, and who you want to become. As teachers we try to create these pockets of opportunity for our students. Yet creating these pockets in increasingly digitised, competitive, and global learning environments raises a series of implications.
We are living in a moment where wealthy technology firms are becoming better and better at using proprietary algorithms to automate ever more aspects of human expression: reading, writing, singing, joking. Through my research, it has become increasingly clear to me that the apparent efficiency of technology in a learning context does not simply come from their algorithmic elegancy and computing power, but is created by moving the pedagogical goalpost for the activity that it automates. In this case, the purpose of learning. For example, if software can make you sound like a good singer at the click of a button, then this will probably be cheaper and easier to do than the alternative, which is to actually teach you how to do it. In the same way, it would be much easier for me (and by extension cheaper for my employer) to only teach students how to write with Large Language Models (LLMs). This is of course important and useful. But the pedagogical purpose and therefore the value of prompting and editing LLM outputs is different from that of carefully crafting your own ideas: learning to write clearly means learning to think clearly. It is a key skill that can help elevate students from knowledge consumers to cutting-edge knowledge producers. As Helen Beetham asks: why is writing “developmental, or how do we make it so? And what kind of people are developed through the writing we ask them to do?” As a bare minimum, I want to help students develop into learned individuals who are able to discern the rhetorical patterns, nudges, framings, and assumptions that are produced by the algorithms they use.
But as an educator, I want more for my students. Rather than debate whether there is room for generative AI at university, the real fault line should instead be whether education must prioritise the becoming of free, autonomous, capable, and productive people: ensuring that students master these new technologies should not come at the expense of higher learning as defined above. Just like there is value in practicing vocal breaks over and over again to achieve mastery, so is there value in learning to read an academic article multiple times and getting under the skin of another’s research project. There is value in carefully crafting an argument, in speaking with other students face-to-face, in finding refuge in a quiet corner in the library. These activities are valuable, not because we can put them on a LinkedIn profile or list them in a job application, but because they change who we are as people, and who we want to become. The magic of these activities’ resides in their fleeting, social, but also introspective nature.
Professor Gourlay unpacks such magic through ephemerality, seclusion, and copresence, which are aspects of academic practice that can make them hard to observe and track. In short, much of academic life should be characterised by seeing fleeting ideas come and go, sitting by yourself with a book, and engaging with peers in physical space with the certainty that the encounter is not recorded. These practices, Gourlay argues, are fugitive. The escape is not from the digital, but from the totalising, ubiquitous, and unfettered network connectivity that constantly infiltrates the learning pockets we create together. Speeding up what must be slowed down, freezing what must be flowing, and valorising what must be priceless.
Such assertions, inevitably, are grounded in normative judgments about the kind of education we should offer. My view on this is simple: learning institutions must both engage with but maintain independence from the society they seek to reproduce and transform. After all, the value of academic life lies in what it is doing to people and communities engaging in it: How it provides students with pockets of humanity that allow them to just be, and in doing so, help them in the process of becoming.
A fully funded PhD position is now available at King’s College London on the project “‘Lost for words’: semantic search in the Find Case Law service of The National Archives”, a Collaborative Doctoral Award received by King’s College London in collaboration with The National Archives and funded by the London Arts & Humanities Partnership (LAHP). This interdisciplinary project is an exciting opportunity to work in natural language processing (particularly computational semantics and information retrieval) applied to legal texts and digital humanities.
About the project: Access to case law is vital for safeguarding the constitutional right of access to justice. It enables members of the public to understand their position when facing litigation and to scrutinise court judgements. Since April 2022, UK court and tribunal decisions are preserved by The National Archives’ Find Case Law service as freely accessible online public records. This project seeks to improve Find Case Law by enhancing it with meaning-sensitive (semantic) search functionality. It will study how individuals without legal training use language to navigate court judgments and it will develop tools to facilitate this navigation. In most digital cultural heritage catalogues, while we can search for words within the metadata describing their records, we cannot search for records based on the meaning of words contained within these records, for example the different words to refer to “knife crime”. Therefore, users’ access to collection is determined by their ability to articulate their information need precisely. Recent advances in natural language processing unlock new possibilities for querying documents via state-of-the-art semantic search. Incorporating such search capabilities in the Find Case Law collection is crucial for democratising access to digital collections, helping expose the social impact of how the law is written.
Skills required
Essential:
Experience with Natural Language Processing research and applied work, including developing new tools.
Interest in working with UK case law for improving access to justice
We invite you to our workshop on ‘the infrastructures of socio-ecological knowing in the city’ that will take place on the 11thJanuary 2024, at King’s College London. The output of this workshop will be proposed as a special issue, which will be submitted in late Spring of 2024. Contributors to the workshop will be encouraged to submit to the issue, understanding the workshop to be a space for developing works in progress, rather than necessarily presenting complete papers.
Please submit your 200 word proposals by the end of 14th December by emailing to gunes.tavmen@kcl.ac.uk
We have funds to cover (at least partially) the travel expenses for Early Career Researchers who otherwise don’t have funding. If you need to obtain visa to travel to the UK, let us know and we’ll expedite the decision making process for your abstract submission.
Overview
When the concept of the smart city emerged, one of its primary promises was to make cities more sustainable. With ubiquitous sensing and real time data flows, city administrators, we were told, would be able to monitor levels of urban pollution, energy use and air quality, which would lead to more efficient and sustainable management of resources. Whilst it is true that technologies, including cameras. sensors, and more recently AI, have been effective at recognizing and tracking environmental impacts, there is also much evidence that highlights the environmental cost of these same technologies, which rely on energy intensive infrastructures (Monserrate, 2022). Moreover, several scholars have observed that the quantifiable logics of data collection on this scale (the datafication of pollution, tree coverage, air quality etc.) does not necessarily lead to meaningful policy changes or straight forward action (see, for example, Gabrys, 2020).
With the recent progress (and hype) in data and AI technologies, some have argued that other ways of knowing the city have been eclipsed by the episteme of data and algorithmic imaginaries, which offer seemingly objective views on urban processes due to their impressive technical capabilities (Mattern, 2017). Following Louise Amoore’s (2020) work, however, we know that AI and Machine Learning techniques can be understood as an aperture – or a perceived opening to new ways of knowing, but also a foreclosing of possible futures into computational and statistical ways of knowing. So, taking the urbanists Brenner and Schmid (2015)’s question: ‘through what categories, methods and cartographies should urban life be understood’?, we instead ask, ‘through what other categories, methods, and technologies could urban ecology be understood?’
We invite papers that are grounded in the reality of how data and AI technologies, and the situated socio-political ways they have become embedded in city governance, have come to shape our understanding of ecological processes in the city, thus moving us away from the imaginaries of smart city techno-utopias. We aim to bring together an understanding of the current state of play, but also to develop future directions for urban-ecological relations that are guided not by today’s focus on datafication and algorithmic processing, but by other ways of knowing the city.
Some questions to guide submissions:
How do sensing and algorithmic technologies shape understandings and perceptions of socio-ecological systems in cities, and how might they foreclosure other ways of knowing?
Which ways of knowing the city are obscured by our society’s focus on the logics of datafication?
How are socio-ecological relations made visible through the lens of digital media? And who are they made visible for?
What role do digital media platforms play in our understanding of socio-ecological ways of knowing the city?
What are the ways of socio-ecological knowing that may be localised, digital or non-digital that might help in working towards environmental justice in the urban environment?
UK universities are an increasingly popular choice for international students, especially those seeking to experience life in multicultural, diverse centres indeed global cities such as London. Nonetheless, getting used to life in large diverse urban settings is not without its challenges. Several studies have reported students facing difficulties ranging from acculturation and socialisation, to datafication and surveillance. To address these challenges, digital media use and information sharing have become critical tools among students to mobilize resources or seize opportunities, but also, to deal with change, uncertainty or risks. The Outsmarted project sought to design an innovative participatory toolkit for understanding how our students met these challenges through and with digital media technologies. Amongst these technologies, understanding the impact of social media and mobility apps was a particular priority, as they have become primary sources for orientation among students, particularly prevalent in global cities such as London, often shaping spatial understandings and knowledge about communications infrastructures, networks and cultures that regulate young people’s life and access to the city’s material and symbolic resources.
In addition to London’s global standing, the city has been at the centre of debates and discourse regarding how digital media technologies are shaping contemporary urban life. It has been a space for experimentation, innovation and dissent surrounding digital models of urbanism – smart, data driven, algorithmic urbanism. For proponents, such models often promise the delivery of new efficiencies, conveniences, and, capacities. For critics, such models position users, or rather city dwellers, as actors devoid of agency or knowledge beyond the machine of urban administration. In The City is not a computer, Shannon Mattern argues that agency and local knowledge should be at the heart of urban intelligence. By studying student experiences in London, the Outsmarted team sought to understand how they make sense of, navigate, and, ultimately become dwellers of a city rapidly transformed by digital technology and smart devices.
The Outsmarted toolkit
Inspired by critical technology, urban and education studies, the Outsmarted project, led by Dr Alevizou, worked with KCL students from Digital Humanities and Liberal Arts to explore their experiences of London in terms of digital culture and communication infrastructures, places and spaces for knowledge, learning and leisure. Deploying a range of participatory pedagogies and qualitative methods, as well as creative and computational, systems thinking, it drew insights into students’ understanding of the ‘digital’ (culture, media, connections, networks and infrastructures and frames surrounding ‘appification’ and AI) through the city and with the city.
Participatory mapping workshops and group discussions, March 2023
This involved a place mapping and ‘asset’ mapping events aiming to reflect on experiences and uses of media, apps and AI tools as well as considerations of resources, needs and obstacles across the following sectors: education and health, space, mobility and culture.
Discussions following the ‘mapping activities’ evoked a sense of participatoryagency, as students contributed experiences as city dwellers and offered reflections surrounding challenges such as information/cognitive overload, mis/dis-information, surveillance as well as privacy concerns and tensions surrounding representation and algorithmic biases. Students reflected that ‘computational thinking’, and the creative, tactile elements of this methodology stimulated discussions that enabled them to collectively cultivate capacities for developing critical digital and AI literacies.
Funding from the Digital Futures Institute enabled us to further reflect upon the existing instruments of the asset mapping toolkit. We used student feedback to refine and reproduce both design and frameworks for discussion that involves a ‘knowledge game’. Deploying gamification, the toolkit can be used as board gameto:
offer new ways to learn about how the digital and the algorithmic is embedded in civic, personal and public domains of life in cities & to reflect on how to make meaningful change
to examine how sociality, subjectivity, and bias, affect become embodied practices within urban communication & digital infrastructures
We believe that deploying creative methods for enabling dialogue alongside conventional focus group and qualitative interviewing methods, allows for building an understanding of ‘knowledge’ or ‘intelligence’ around/through the ‘digital city’, and for, creating platforms for learning rights and for cultivating civic capacities. Living well with technology means engaging with technical processes and systems; recognizing that they are not neutral; and unpacking their agencies and their entanglement with how they are produced and used. It means awareness and intervention.
The Outsmarted creative methods toolkit
Future Plans
We will use this toolkit as a further learning resource, based on research- informed collaborative pedagogies for DDH’s modules in digital culture and media. A series of workshops will be advertised in the 2023-24 academic year for those wishing to learn more about the project and how to integrate the toolkit into their research and teaching and with a view to developing critical digital and AI literacies.
The toolkit can be accessed online and cited as: Alevizou Giota, Duggan, Michael & Photini Vrikki. (2023). Outsmarted? A Creative Methods Toolkit for reflecting on experiences of London as a Digital City. Zenodo. https://doi.org/10.5281/zenodo.8341806
The article explores testing situations – moments in which it is no longer possible to go on in the usual way – across scales during the COVID-19 pandemic through interpretive querying and sub-setting of Twitter data (“data teasing”), together with situational image analysis.
The full text is available open access here. Further details and links can be found at this project page. The abstract and reference are copied below.
How was testing—and not testing—for coronavirus articulated as a testing situation on social media in the Spring of 2020? Our study examines everyday situations of Covid-19 testing by analyzing a large corpus of Twitter data collected during the first 2 months of the pandemic. Adopting a sociological definition of testing situations, as moments in which it is no longer possible to go on in the usual way, we show how social media analysis can be used to surface a range of such situations across scales, from the individual to the societal. Practicing a form of large-scale data exploration we call “interpretative querying” within the framework of situational analysis, we delineated two types of coronavirus testing situations: those involving locations of testing and those involving relations. Using lexicon analysis and composite image analysis, we then determined what composes the two types of testing situations on Twitter during the relevant period. Our analysis shows that contrary to the focus on individual responsibility in UK government discourse on Covid-19 testing, English-language Twitter reporting on coronavirus testing at the time thematized collective relations. By a variety of means, including in-memoriam portraits and infographics, this discourse rendered explicit challenges to societal relations and arrangements arising from situations of testing and not testing for Covid-19 and highlighted the multifaceted ways in which situations of corona testing amplified asymmetrical distributions of harms and benefits between different social groupings, and between citizens and state, during the first months of the pandemic.
Marres, N., Colombo, G., Bounegru, L., Gray, J. W. Y., Gerlitz, C., & Tripp, J. (2023). Testing and Not Testing for Coronavirus on Twitter: Surfacing Testing Situations Across Scales With Interpretative Methods. Social Media + Society, 9(3). https://doi.org/10.1177/20563051231196538
