|Modeling news recommender systems’ conditional effects on selective exposure: evidence from two online experiments Journal Article |
In: Journal of Communication , 2023.
Under which conditions do news recommender systems (NRSs) amplify or reduce selective exposure? I provide the Recommender Influenced Selective Exposure framework, which aims to enable researchers to model and study the conditional effects of NRSs on selective exposure. I empirically test this framework by studying user behavior on a news site where the choice environment is designed to systematically influence selective exposure. Through two preregistered online experiments that simulate different NRSs and unobtrusively log user behavior, I contribute empirical evidence that an NRS can increase or decrease the chance that selective exposure occurs, depending on what the NRS is designed to achieve. These insights have implications for ongoing scholarly debates on the democratic impact of NRSs.
|The future technologies of journalism Conference |
Bjørnar Tessem; Are Tverberg; Njål Borch
The practice of journalism has undergone many changes in the last few years, with changes in technology being the
main driver of these changes. We present a future study where we aim to get an understanding of what technologies
will become important for the journalist and further change the journalist’s workplace. The new technological
solutions will have to be implemented in the media houses’ information systems, and knowledge about what
technologies will have the greatest impact will influence IS strategies in the media house. In the study we
interviewed 16 experts on how they envision the future technologies of the journalist. We analyzed the interviews
with a qualitative research approach. Our analysis shows that technologies for multi-platform news production,
automated news content generation, cloud services for flexible production, content search, and content verification
are the most important in terms of needs and competitiveness.
|Understanding How News Recommender Systems Influence Selective Exposure Conference |
Khadiga Seddik; Erik Knudsen; Damian Trilling; Christoph Trattner
Association for Computing Machinery (ACM) RecSys ’23, 2023.
|Topical Preference Trumps Other Features in News Recommendation: A Conjoint Analysis on a Representative Sample from Norway Conference |
Erik Knudsen; Alain D. Starke; Christoph Trattner
Association for Computing Machinery (ACM) RecSys ’23, 2023.
|Evaluating The Effects of Calibrated Popularity Bias Mitigation: A Field Study Conference |
Anastasiia Klimashevskaia; Mehdi Elahi; Dietmar Jannach; Lars Skjærven; Astrid Tessem; Christoph Trattner
Association for Computing Machinery (ACM) RecSys ’23, 2023.
|The Interplay between Food Knowledge, Nudges, and Preference Elicitation Methods Determines the Evaluation of a Recipe Recommender System Conference |
Ayoub El Majjodi; Alain D. Starke; Christoph Trattner
Association for Computing Machinery (ACM) RecSys ’23, 2023.
|Using Visual and Linguistic Framing to Support Sustainable Decisions in an Online Store Conference |
Alain Starke; Kimia Emami; Andrea Makarová; Bruce Ferwerda
Association for Computing Machinery (ACM) RecSys ’23,, 2023.
|A software reference architecture for journalistic knowledge Journal Article |
Marc Gallofré Ocaña; Andreas L. Opdahl
In: Knowledge-based Systems, vol. 276, 2023.
Newsrooms and journalists today rely on many different artificial-intelligence, big-data and knowledgebased systems to support efficient and high-quality journalism. However, making the different systems
work together remains a challenge, calling for new unified journalistic knowledge platforms. A software
reference architecture for journalistic knowledge platforms could help news organisations by capturing
tried-and-tested best practices and providing a generic blueprint for how their IT infrastructure should
evolve. To the best of our knowledge, no suitable architecture has been proposed in the literature.
Therefore, this article proposes a software reference architecture for integrating artificial intelligence
and knowledge bases to support journalists and newsrooms. The design of the proposed architecture
is grounded on the research literature and on our experiences with developing a series of prototypes
in collaboration with industry. Our aim is to make it easier for news organisations to evolve their
existing independent systems for news production towards integrated knowledge platforms and to
direct further research. Because journalists and newsrooms are early adopters of integrated knowledge
platforms, our proposal can hopefully also inform architectures in other domains with similar needs.
|Construction of a relevance knowledge graph with application to the LOCAL news angle Inproceedings |
Bjørnar Tessem; Marc Gallofré Ocaña; Andreas L. Opdahl
In: CEUR Workshop Proceedings (CEUR-WS.org) , 2023.
News angles are approaches to journalism content often used to provide a way to present a new report
from an event. One particular type of news angle is the LOCAL news angle where a local news outlet
focuses on an event by emphasising a local connection. Knowledge graphs are most often used to
represent knowledge about a particular entity in the form of relationships to other entities. In this paper
we see how we can extract a knowledge sub graph containing entities and relevant relationships that are
connected to the locality of a news outlet. The purpose of this graph is to use it for automated journalism
or as an aid for the journalist to find local connections to an event, as well as how the local connection
relate to the event. We call such a graph a relevance knowledge graph. An algorithm for extracting such
a graph from a linked data source like DBpedia is presented and examples of the use of a relevance graph
in a LOCAL news angle context are provided.
|NorBench – A Benchmark for Norwegian Language Models Conference |
David Samuel; Andrey Kutuzov; Samia Touileb; Erik Velldal; Lilja Øvrelid; Egil Rønningstad; Elina Sigdel; Anna Palatkina
We present NorBench: a streamlined suite of NLP tasks and probes for evaluating Norwegian language models (LMs) on standardized data splits and evaluation metrics. We also introduce a range of new Norwegian language models (both encoder and encoder-decoder based). Finally, we compare and analyze their performance, along with other existing LMs, across the different benchmark tests of NorBench.
|Identifying Token-Level Dialectal Features in Social Media Conference |
Jeremy Barnes, Samia Touileb, Petter Mæhlum; Pierre Lison
Dialectal variation is present in many human languages and is attracting a growing interest in NLP. Most previous work concentrated on either (1) classifying dialectal varieties at the document or sentence level or (2) performing standard NLP tasks on dialectal data. In this paper, we propose the novel task of token-level dialectal feature prediction. We present a set of fine-grained annotation guidelines for Norwegian dialects, expand a corpus of dialectal tweets, and manually annotate them using the introduced guidelines. Furthermore, to evaluate the learnability of our task, we conduct labeling experiments using a collection of baselines, weakly supervised and supervised sequence labeling models. The obtained results show that, despite the difficulty of the task and the scarcity of training data, many dialectal features can be predicted with reasonably high accuracy.
|Automated Claim Detection for Fact-checking: A Case Study using Norwegian Pre-trained Language Models Conference |
Ghazaal Sheikhi; Samia Touileb; Sohail Ahmed Khan
We investigate to what extent pre-trained language models can be used for automated claim detection for fact-checking in a low resource setting. We explore this idea by fine-tuning four Norwegian pre-trained language models to perform the binary classification task of determining if a claim should be discarded or upheld to be further processed by human fact-checkers. We conduct a set of experiments to compare the performance of the language models, and provide a simple baseline model using SVM with tf-idf features. Since we are focusing on claim detection, the recall score for the upheld class is to be emphasized over other performance measures. Our experiments indicate that the language models are superior to the baseline system in terms of F1, while the baseline model results in the highest precision. However, the two Norwegian models, NorBERT2 and NB-BERT_large, give respectively superior F1 and recall values. We argue that large language models could be successfully employed to solve the automated claim detection problem. The choice of the model depends on the desired end-goal. Moreover, our error analysis shows that language models are generally less sensitive to the changes in claim length and source than the SVM model.
|Measuring Normative and Descriptive Biases in Language Models Using Census Data Conference |
Samia Touileb; Lilja Øvrelid; Erik Velldal
We investigate in this paper how distributions of occupations with respect to gender is reflected
in pre-trained language models. Such distributions are not always aligned to normative ideals, nor do they necessarily reflect a descriptive assessment of reality. In this paper, we introduce an approach for measuring to what degree pre-trained language models are aligned to normative and descriptive occupational distributions. To this end, we use official demographic information about gender–occupation distributions provided by the national statistics agencies of France, Norway, United Kingdom, and the United States. We manually generate template-based sentences combining gendered pronouns and nouns with occupations,
and subsequently probe a selection of ten language models covering the English, French, and Norwegian languages. The scoring system we introduce in this work is language independent, and can be used on any combination of
template-based sentences, occupations, and languages. The approach could also be extended to other dimensions of national census data and other demographic variables.
|Trustworthy Journalism Through AI Journal Article |
Andreas L. Opdahl; Bjørnar Tessem; Duc-Tien Dang-Nguyen; Enrico Motta; Vinay Setty; Eivind Throndsen; Are Tverberg; Christoph Trattner
In: Data & Knowledge Engineering (DKE), Elsevier, 2023.
Quality journalism has become more important than ever due to the need for quality and trustworthy media outlets that can provide accurate information to the public and help to address and counterbalance the wide and rapid spread of disinformation. At the same time, quality journalism is under pressure due to loss of revenue and competition from alternative information providers. This vision paper discusses how recent advances in Artificial Intelligence (AI), and in Machine Learning (ML) in particular, can be harnessed to support efficient production of high-quality journalism. From a news consumer perspective, the key parameter here concerns the degree of trust that is engendered by quality news production. For this reason, the paper will discuss how AI techniques can be applied to all aspects of news, at all stages of its production cycle, to increase trust.
|Towards Attitudinal Change in News Recommender Systems: A Pilot Study on Climate Change Workshop |
Jia Hua Jeng; Alain D. Starke; Christoph Trattner
Personalized recommender systems facilitate decision-making in various domains by presenting content closely aligned with users’ preferences.
However, personalization can lead to unintended consequences. In news, selective information exposure and consumption might amplify
polarization, as users are empowered to seek out information that is in line with their own attitudes and viewpoints. However, personalization in
terms of algorithmic content and persuasive technology could also help to narrow the gap between polarized user attitudes and news consumption
patterns. This paper presents a pilot study on climate change news. We examined the relation between users’ level of environmental concern, their preferences
for news articles, and news article content. We aimed to capture a news article’s viewpoint through sentiment analysis. Users (N = 180)
were asked to read and evaluate 10 news articles from the Washington Post. We found a positive correlation between users’ level of environmental
concern and whether they liked the article. In contrast, no significant correlation was found between sentiment and environmental concern.
We argue why a different type of news article analysis than sentiment is needed. Finally, we present our research agenda on how persuasive technology
might help to support more exploration of news article viewpoints in the future.
|The Burden of Subscribing: How Young People Experience Digital News Subscriptions Journal Article |
Marianne Borchgrevink-Brækhus; Hallvard Moe
In: Journalism Studies , 2023.
This paper analyzes how young non-paying news users experience digital news subscriptions in Norway. As news organizations face
declining advertising revenues, digital subscriptions are considered the sustainable financial strategy of the future, with young people a particularly challenging group to convert. We analyze the experiences of young adults who do not pay for news and identify three key dimensions to why they do not subscribe:
lack of exclusivity, subscriptions as too time-consuming, and unattractive payment models. We also detail how the informants maneuver around paywalls, and we highlight “multiperspectivism” as an overarching concern guiding the informants’ preferences. Empirically, the paper furthers our understanding of
the challenges facing business models for journalism, especially problems with long-term, provider-specific subscriptions. Methodologically, we demonstrate how a combination of recurring interviews and a media diary matching a subscription test period yields a deeper analysis of motivations for, and
experiences with, news use. Theoretically, the paper shows how approaching news through users’ experiences can provide insights not just into what users appreciate from news but also into where they consider there is a lack of value.
|Healthiness and environmental impact of dinner recipes vary widely across developed countries Journal Article |
Aslaug Angelsen; Alain D. Starke; Christoph Trattner
In: Nature Food , 2023.
Contrary to food ingredients, little is known about recipes’ healthiness or environmental impact. Here we examine 600 dinner recipes from Norway, the UK and the USA retrieved from cookbooks and the Internet. Recipe healthiness was assessed by adherence to dietary guidelines and aggregate health indicators based on front-of-pack nutrient labels, while environmental impact was assessed through greenhouse gas emissions and land use. Our results reveal that recipe healthiness strongly depends on the healthiness indicator used, with more than 70% of the recipes being classified as healthy for at least one front-of-pack label, but less than 1% comply with all dietary guidelines. All healthiness indicators correlated positively with each other and negatively with environmental impact. Recipes from the USA, found to use more red meat, have a higher environmental impact than those from Norway and the UK.
|Fairness in automated data journalism systems Journal Article |
Bahareh Fatemi; Fazle Rabbi; Bjørnar Tessem
In: NIKT: Norsk IKT-konferanse for forskning og utdanning, 2023.
Automated data journalism is an application of computing and artificial intelligence (AI) that aims to create stories from raw data, possibly in a variety of formats (such as visuals or text). Conventionally, a variety of methodologies and tools, including statistical software packages and data visualization tools have been used to generate stories from raw data. Artificial intelligence, and particularly machine learning techniques have recently been introduced because they can handle more complex data and scale more easily to larger datasets. However, AI techniques may raise a number of ethical concerns such as an unfair presentation which typically occurs due to bias. Stories that contains unfair presentation could be destructive at individual and societal levels; they could also damage the reputation of news providers. In this paper we study an existing framework of automated journalism and enhance the framework to make it aware of fairness concern. We present various steps of the framework where bias enters into the production of a story and address the causes and effects of different types of biases.
|How Rally-Round-the-Flag Effects Shape Trust in the News Media: Evidence from Panel Waves before and during the COVID-19 Pandemic Crisis Journal Article |
Erik Knudsen; Åsta Dyrnes Nordø; Magnus Hoem Iversen
In: Political Communication, 2023.
In this study, we extend the literature on the rally ‘round the flag phenomenon, that is, that international crises tend to cause an increase in citizens’ approval of political institutions. We advance this literature and highlight its relevance for political communication research in three ways: 1) by theorizing and empirically testing two arguments for why rally effects should extend to trust in the news media on the institutional level, 2) by providing empirical evidence on how rally effects on trust in the media develop over time during an international crisis, and 3) by theorizing and testing the conditions under which rally effects on media trust are more likely to occur by studying heterogeneous effects. Through a panel design with a pre-crisis baseline of Norwegian citizens’ trust in news media, we find evidence to suggest that the compound effect of the COVID-19 pandemic crisis caused a long-lasting increase in trust in the news media in Norway, and that the degree of increase varied by citizens’ education and whether they belonged to a “high-risk” group. We also provide evidence to suggest that rally effects on news media trust are contingent on how important the news media is as a source of information about the crisis and the “trust nexus” between media trust and political trust. These insights extend our current understanding of how times of crisis affect trust in the news media.
|Visual User-Generated Content Verification in Journalism: An Overview Journal Article |
Sohail Ahmed Khan; Ghazaal Sheikhi; Andreas L. Opdahl; Fazle Rabbi; Sergej Stoppel; Christoph Trattner; Duc-Tien Dang-Nguyen
In: IEEE Access, 2023.
Over the past few years, social media has become an indispensable part of the news generation and dissemination cycle on the global stage. These digital channels along with the easy-to-use editing tools have unfortunately created a medium for spreading mis-/disinformation containing visual content. Media practitioners and fact-checkers continue to struggle with scrutinising and debunking visual user-generated content (UGC) quickly and thoroughly as verification of visual content requires a high level of expertise and could be exceedingly complex amid the existing computational tools employed in newsrooms. The aim of this study is to present a forward-looking perspective on how visual UGC verification in journalism can be transformed by multimedia forensics research. We elaborate on a comprehensive overview of the five elements of the UGC verification and propose multimedia forensics as the sixth element. In addition, different types of visual content forgeries and detection approaches proposed by the computer science research community are explained. Finally, a mapping of the available verification tools media practitioners rely on is created along with their limitations and future research directions to gain the confidence of media professionals in using multimedia forensics tools in their day-to-day routine.
|Monitoring the infection rate: Explaining the meaning of metrics in pandemic news experiences Journal Article |
John Magnus Ragnhildson Dahl; Brita Ytre-Arne
In: Journalism , 2023.
The COVID-19 pandemic has brought forward questions of what citizens need and want from journalism in a global crisis. In this article, we analyse one particular aspect of pandemic news experiences: Preoccupation with monitoring metrics for COVID-19 infection cases, hospitalisations, and deaths, widely disseminated through journalistic news outlets. We ask why close monitoring of such metrics appeared meaningful to news users, and what these experiences can tell us about the role of journalism in the pandemic information environment. Our analysis draws on qualitative research conducted in Norway in 2020, finding users particularly devoted to monitoring metrics, both in early lockdown and during the second wave of COVID-19. To contextualize our findings, we draw on scholarship on emotional responses to data in the everyday, and on the social role of journalism. We argue that monitoring of infection rates is an expression of trust in the media as a provider of factual information, also expressed by those who are cynical towards other aspects of journalism, and we conceptualise this monitoring practice as a coping strategy to deal with the pandemic as an unknown and uncontrollable threat, involving difficult emotions of uncertainty and fear.
|Semantic Knowledge Graphs for the News: A Review Journal Article |
Andreas L. Opdahl; Tareq Al-Moslmi; Duc-Tien Dang-Nguyen; Marc Gallofré Ocaña
In: ACM Computing Surveys, vol. 55, iss. 7, pp. 1-38, 2022.
ICT platforms for news production, distribution, and consumption must exploit the ever-growing availability of digital data. These data originate from different sources and in different formats; they arrive at different velocities and in different volumes. Semantic knowledge graphs (KGs) is an established technique for integrating such heterogeneous information. It is therefore well-aligned with the needs of news producers and distributors, and it is likely to become increasingly important for the news industry. This article reviews the research on using semantic knowledge graphs for production, distribution, and consumption of news. The purpose is to present an overview of the field; to investigate what it means; and to suggest opportunities and needs for further research and development.
|Measuring Harmful Representations in Scandinavian Language Models Conference |
Samia Touileb; Debora Nozza
Scandinavian countries are perceived as rolemodels when it comes to gender equality. With the advent of pre-trained language models and their widespread usage, we investigate to what extent gender-based harmful and toxic content exist in selected Scandinavian language models. We examine nine models, covering Danish, Swedish, and Norwegian, by manually creating template-based sentences and probing
the models for completion. We evaluate the completions using two methods for measuring harmful and toxic completions and provide a thorough analysis of the results. We show that Scandinavian pre-trained language models contain harmful and gender-based stereotypes with similar values across all languages.
This finding goes against the general expectations related to gender equality in Scandinavian countries and shows the possible problematic outcomes of using such models in real world settings.
|Research directions in recommender systems for health and well-being Journal Article |
Hanna Hauptmann; Alan Said; Christoph Trattner
In: User Modeling and User-Adapted Interaction Journal , 2022.
Recommender systems have been put to use in the entertainment and e-commerce domains for decades, and in these decades, recommender systems have grown and matured into reliable and ubiquitous systems in today’s digital landscape. Relying on this maturity, the application of recommender systems for health and well-being has seen a rise in recent years, paving the way for tailored and personalized systems that support caretakers, caregivers, and other users in the health domain. In this introduction, we give a brief overview of the stakes, the requirements, and the possibilities that recommender systems for health and well-being bring.
|Journalism Education for Datafied Society: Fostering Data (infrastructural) literacy Journal Article |
In: Social Science Research Network (SSRN), 2022.
This paper argues for introducing data (infrastructural) literacy as a potential response of journalism education to the challenge of growing datafication in the media industry. It begins with pointing to the vocational orientation and centrality of technology in journalism education. Moving on to datafication as the most recent technological revolution and respective changes it is bringing in the news work. After brief mapping out of the existing responses of journalism education to the challenges posed by datafication, arguments in favor of introducing data infrastructural literacy in curricula of journalism education are presented. Arguments raised in this paper are based on the review of the previous literature as well as the insights gained from the in-depth semi-structured interviews with data analysts from the most influential media organizations in Norway.
|Annotating Norwegian language varieties on Twitter for Part-of-speech Workshop |
Petter Mæhlum; Andre Kåsen; Samia Touileb; Jeremy Barnes
Norwegian Twitter data poses an interesting challenge for Natural Language Processing (NLP) tasks. These texts are difficult for models trained on standardized text in one of the two Norwegian written forms (Bokmål and Nynorsk), as they contain both the typical variation of social media text, as well as a large amount of dialectal variety. In this paper we present a novel Norwegian Twitter dataset annotated with POS-tags. We show that models trained on Universal Dependency (UD) data perform worse when evaluated against this dataset, and that models trained on Bokmål generally perform better than those trained on Nynorsk. We also see that performance on dialectal tweets is comparable to the written standards for some models. Finally we perform a detailed analysis of the errors that models commonly make on this data.
|Datafication Media and Democracy: Audience Analytics and Metrics in the Norwegian Media Organizations Journal Article |
In: Social Science Research Network (SSRN), 2022.
The audience analytics and metrics are becoming integral part of news work, with growing importance for defining strategic goals, editorial decision making and everyday journalistic practice. While media organizations are embracing data informed news making, academic research is critical about many aspects of analytics use. This paper provides overview of the use of audience metrics and analytics in most influential media organizations in Norway with aim to showcase how analytics are utilized in one of the most solid media industries in Europe. Results presented in this paper are based on in-depth semi-structured interviews with data analysts, editors and reporters from broadcasting and publishing media organizations.
|360TourGuiding: Towards Virtual Reality Training for Tour Guiding Conference |
Duy-Nam Ly; Thanh-Thai La; Khanh-Duy Le; Cuong Nguyen; Morten Fjeld; Thanh Ngoc-Dat Tran; Minh-Triet Tran
360TourGuiding: Towards Virtual Reality Training for Tour Guiding, 2022.
Tour guiding plays an important role in turning sightseeing tours into memorable experiences. Tour guides, especially inexperienced ones, must practice intensively to perfect their craft. It is key that guides acquire knowledge about sights, in-situ presentation skills, and perfection ability to interact with and engage tourists. Therefore, tour-guide education requires on-site training at the place of interest including live tourist audiences. However, for modest budgets, such setups are costly and tourism students have to practice tour guiding at home or in simulated class-room setups. It has become a challenge for students to adequately prepare themselves for jobs in terms of relevant knowledge and skills. To tackle this problem, we propose 360TourGuiding, a VR system enabling its users to practice tour guiding with 360 travel videos plus the attendance of remote audiences participating through their mobile and personal device. This paper reports on the concept, on our design, current implementation, and on a pilot study with the current 360TourGuiding prototype. Based on qualitative feedback gained through the pilot study, we discuss possible system improvements, future system updates, and plans for empirical evaluation.
|Examining Choice Overload across Single-list and Multi-list User Interfaces Inproceedings Forthcoming|
Alain Starke, Justyna Sedkowska, Mihir Chouhan; Bruce Ferwerda
Recommender systems are prone to triggering choice overload among users due to the typically large set sizes. Various applications have been developed that aim to overcome this through interface design, notably by so-called multi-list recommender systems. However, to what extent such user interface design actually reduces choice overload compared to single-list interfaces has yet to be examined. In a user study, we compared three common user interfaces (UIs) in the context of recipe recommendation: a single-list UI, a grid UI and a multi-list UI. Whereas earlier studies found differences in choice difficulty and choice satisfaction across grid-based and multi-list recommender interfaces, we observed no such differences, as the explanations were possibly not sufficiently helpful. Instead, we found that grid-based UIs and multi-list UIs had a higher perceived ease of use than a single-list UI, which in turn reduced choice difficulty. The benefits of such interfaces may, thus, lie in the organization of the UI, at least in the recipe domain.
|"Det er ikke plass til alt på internett": algoritmestyrte forsider og redaksjonelle vurderinger Journal Article |
In: Norsk Medietidsskrif, 2022.
Denne artikkelen er en analyse av overgangen til algoritmestyrte forsider i to store norske nettaviser. I 2019 innførte flere norske regionaviser i Schibsted-konsernet algoritmestyrte nettavisforsider. Mens sakene på forsidene tidligere ble manuelt styrt og rangert av frontsjefene, er forsidene nå i stor grad automatisert ut fra bruksdata. Algoritmer automatiserer deler av forsiden basert på salg, trafikk og nyhetsverdi, der formålet, ifølge utviklerne, er å frigjøre tid til viktigere journalistiske oppgaver og gi leserne flere relevante saker. Omleggingen har imidlertid skapt debatt og bekymring for at algoritmene kan bidra til å personalisere nyhetstilbudet og svekke opplevelsen av en felles dagsor- den. Gjennom kvalitative intervjuer med utviklere, frontsjefer og journalister i to av Norges største abonnementsavi- ser, Aftenposten og Bergens Tidende, undersøker artikkelen hvordan algoritmene systematiserer nyhetsinnholdet, og hva innføringen av algoritmene betyr for tilgangen på nyheter. Selv om forsidene til en viss grad personaliseres, viser funnene at algoritmenes nåværende utforming ikke har de samme selvforsterkende mekanismene som ofte forbindes med sosiale medier. Samtidig ser redaksjonene for seg at algoritmene i fremtiden kan bidra til likere informasjons- og nyhetstilgang blant ulike brukergrupper.
This article is an analysis of the transition to algorithm-driven front pages in two Norwegian online newspapers. In 2019, Norwegian media group Schibsted implemented front page algorithms in several of their online newspapers. While the news stories previously were ranked manually by editors, the front pages are now partly automated based on sales, clicks, and news value. According to web developers, the purpose is to free up time for more critical jour- nalistic tasks and provide their readers with more relevant content. Nevertheless, the implementation of the technol- ogy has caused concern. Critics fear that algorithms will personalize the news and weaken the experience of a shared public agenda. Drawing on qualitative interviews with web developers, front page editors, and journalists in the Nor- wegian newspapers Aftenposten and Bergens Tidende, the article examines how the front page algorithms systemize and rank news content. Although the front pages are moderately personalized, the findings show that the current design of the algorithms does not entail the self-reinforcing mechanisms associated with social media. Furthermore, news professionals imply that algorithmic curation can become essential to ensure equal access to news among dif- ferent user groups in the future.
|Unifying Recommender Systems and Conversational User Interfaces Book Chapter |
Alan Dominique Starke; Minha Lee
In: Chapter 7, pp. 71-77, Association for Computing Machinery, 2022, ISBN: 978-1-4503-9739-1.
|Threshold prediction for detecting rare positive samples using a meta-learner Journal Article |
Hossein Ghaderi Zefrehi; Ghazaal Sheikhi; Hakan Altınçay
Threshold-moving is one of the several techniques employed in correcting the bias of binary classifiers towards the majority class. In this approach, the decision threshold is adjusted to detect the minority class at the cost of increased misclassification of the majority. In practice, selecting a good threshold using cross-validation on the training data is not feasible in some problems since there are only a few minority samples. In this study, building a meta-learner for threshold prediction to tackle the threshold estimation problem in the case of rare positive samples is addressed. Novel meta-features are suggested to quantify the imbalance characteristics of the data sets and the patterns among the prediction scores. A random forest-based threshold prediction model is constructed using these meta-features extracted from the score space of external data. The models obtained are then employed to estimate the optimal thresholds for previously unseen datasets. The random forest-based meta-learner that employs implicitly selected subset of the proposed meta-features and encodes information from multiple external sources in the form of different trees is evaluated by using 52 imbalanced datasets. In the first set of experiments, the best-fitting thresholds are computed for SVM and logistic regression classifiers that are trained using the original imbalanced training sets. The experiments are repeated by using ensembles of multiple learners, each trained using a different balanced data set. It is observed that the proposed approach provides better F-score when compared to alternative threshold-moving and balancing techniques.
|Hybrid Transformer Network for Deepfake Detection Conference |
Sohail Ahmed Khan; Duc-Tien Dang-Nguyen
Hybrid Transformer Network for Deepfake Detection, 2022.
|Playful recognition: Television comedy and the politics of mediated recognition Journal Article |
John Magnus Dahl; Torgeir Uberg Nærland
In: Communication. The European Journal of Communication Research. , 2022.
This article explores how media content may facilitate processes of recognition through playfulness and comedy. Mediated recognition is typically understood as a matter of respectful and positive representation of subaltern groups and in terms of struggles for visibility and dignity. Yet at the same time, the media address audiences in much less deferential ways that are nonetheless consequential to processes of recognition: by means of playfulness, subversion, and irreverence. This article introduces the concept of ‘playful recognition’ to account for the contradictory ways in which humor can incite recognition. The article empirically illustrates this concept drawing upon a case study of Svart Humor – a comedy show aired in Norway. On the one hand, this article explores an important yet neglected dimension of mediated recognition, on the other, it introduces a recognition perspective to the study of televised comedy.
|Nudging Towards Health? Examining the Merits of Nutrition Labels and Personalization in a Recipe Recommender System Conference |
Ayoub El Majjodi; Alain D. Starke; Christoph Trattner
Nudging Towards Health? Examining the Merits of Nutrition Labels and Personalization in a Recipe Recommender System, 2022.
Food recommender systems show personalized recipes to users based on content liked previously. Despite their potential, often recommended (popular) recipes in previous studies have turned out to be unhealthy, negatively contributing to prevalent obesity problems worldwide. Changing how foods are presented through digital nudges might help, but these are usually examined in non-personalized contexts, such as a brick-and-mortar supermarket. This study seeks to support healthy food choices in a personalized interface by adding front-of-package nutrition labels to recipes in a food recommender system. After performing an offline evaluation, we conducted an online study (N = 600) with six different recommender interfaces, based on a 2 (non-personalized vs. personalized recipe advice) x 3 (No Label, Multiple Traffic Light, Nutri-Score) between-subjects design. We found that recipe choices made in the non-personalized scenario were healthier, while the use of nutrition labels (our digital nudge) reduced choice difficulty when the content was personalized.
|Occupational Biases in Norwegian and Multilingual Language Models Workshop |
Samia Touileb; Lilja Øvrelid; Erik Velldal
In this paper we explore how a demographic distribution of occupations, along gender dimensions, is reflected in pre-trained language models. We give a descriptive assessment of the distribution of occupations, and investigate to what extent these are reflected in four Norwegian and two multilingual models. To this end, we introduce a set of simple bias probes, and perform five different tasks combining gendered pronouns, first names, and a set of occupations from the Norwegian statistics bureau. We show that language specific models obtain more accurate results, and are much closer to the real-world distribution of clearly gendered occupations. However, we see that none of the models have correct representations of the occupations that are demographically balanced between genders. We also discuss the importance of the training data on which the models were trained on, and argue that template-based bias probes can sometimes be fragile, and a simple alteration in a template can change a model’s behavior.
|NorDiaChange: Diachronic Semantic Change Dataset for Norwegian Book Chapter |
Andrei Kutuzov; Samia Touileb; Petter Mæhlum; Tita Enstad; Alexandra Witteman
In: pp. 2563-2572, European Language Resources Association NVI-nivå 1, 2022, ISBN: 979-10-95546-72-6.
We describe NorDiaChange: the first diachronic semantic change dataset for Norwegian. NorDiaChange comprises two novel subsets, covering about 80 Norwegian nouns manually annotated with graded semantic change over time. Both datasets follow the same annotation procedure and can be used interchangeably as train and test splits for each other. NorDiaChange covers the time periods related to pre- and post-war events, oil and gas discovery in Norway, and technological developments. The annotation was done using the DURel framework and two large historical Norwegian corpora. NorDiaChange is published in full under a permissive licence, complete with raw annotation data and inferred diachronic word usage graphs (DWUGs).
|Considering Temporal Aspects in Recommender Systems: A Survey Journal Article |
Veronica Bogina; Tsvi Kuflik; Dietmar Jannach; Maria Bielikova; Michal Kompan; Christoph Trattner
In: UMUAI journal, 2022.
Tags: New| |
The widespread use of temporal aspects in user modeling indicates their importance, and their consideration showed to be highly effective in var- ious domains related to user modeling, especially in recommender systems. Still, past and ongoing research, spread over several decades, provided multi- ple ad-hoc solutions, but no common understanding of the issue. There is no standardization and there is often little commonality in considering tempo- ral aspects in different applications. This may ultimately lead to the problem that application developers define ad-hoc solutions for their problems at hand, sometimes missing or neglecting aspects that proved to be effective in similar cases. Therefore, a comprehensive survey of the consideration of temporal as- pects in recommender systems is required. In this work, we provide an overview of various time-related aspects, categorize existing research, present a tempo- ral abstraction and point to gaps that require future research. We anticipate this survey will become a reference point for researchers and practitioners alike when considering the potential application of temporal aspects in their personalized applications.
|Deep Learning to Encourage Citizen Involvement in Local Journalism Book Chapter |
Bjørnar Tessem; Lars Nyre; Paul Mulholland
In: Mari K. Niemi Ville J. E. Manninen, Anthony Ridge-Newman (Ed.): Chapter 3, pp. 211-226, Palgrave Macmillan Cham, 2022.
Tags: New| |
We discuss the potential of a mobile app for news tips to local newspapers to be augmented with artificial intelligence. It can be designed to encourage deliberative, consensus-oriented contributions from citizens. We presume that such an app will generate news stories from multi-modal data in the form of photos, videos, text elements, location information, and the identity of the contributor. Three scenarios are presented to show how image recognition, natural language processing, narrative construction, and other AI technologies can be applied. The scenarios address three interrelated challenges for local journalism. First, text and photos in tips are often of low quality for journalism purposes. Second, peer-to-peer dialogue about local news takes place in social media instead of in the newspaper. Third, readers lack news literacy and are prone to confrontational debates and trolling. We show how advances in deep learning technology makes it possible to propose solutions to these problems.
A Collaborative System of Flying and Ground Robots with Universal Physical Coupling Interface (PCI), and the Potential Interactive Applications Conference
Ziming Wang; Ziyi Hu; Yemao Man; Morten Fjeld
A Collaborative System of Flying and Ground Robots with Universal Physical Coupling Interface (PCI), and the Potential Interactive Applications, 2022.
Flying and ground robots complement each other in terms of their advantages and disadvantages. We propose a collaborative system combining flying and ground robots, using a universal physical coupling interface (PCI) that allows for momentary connections and disconnections between multiple robots/devices. The proposed system may better utilize the complementary advantages of both flying and ground robots. We also describe various potential scenarios where such a system could be of benefit to interact with humans - namely, remote field works and rescue missions, transportation, healthcare, and education. Finally, we discuss the opportunities and challenges of such systems and consider deeper questions which should be studied in future work.
RedirectedDoors: Redirection While Opening Doors in Virtual Reality Conference
Morten Fjeld; Yukai Hoshikawa; Kazuyuki Fujita; Kazuki Takashima; Yoshifumi Kitamura
RedirectedDoors: Redirection While Opening Doors in Virtual Reality., 2022.
We propose RedirectedDoors, a novel technique for redirection in VR focused on door-opening behavior. This technique manipulates the user's walking direction by rotating the entire virtual environment at a certain angular ratio of the door being opened, while the virtual door's position is kept unmanipulated to ensure door-opening realism. Results of a user study using two types of door-opening interfaces (with and without a passive haptic prop) revealed that the estimated detection thresholds generally showed a higher space efficiency of redirection. Following the results, we derived usage guidelines for our technique that provide lower noticeability and higher acceptability.
|Developing and Evaluating a University Recommender System Journal Article |
Mehdi Elahi; Alain D. Starke; Nabil El Ioini; Anna Alexander Lambrix; Christoph Trattner
In: Frontiers in Artificial Intelligence , 2022.
A challenge for many young adults is to find the right institution to follow higher education. Global university rankings are a commonly used, but inefficient tool, for they do not consider a person's preferences and needs. For example, some persons pursue prestige in their higher education, while others prefer proximity. This paper develops and evaluates a university recommender system, eliciting user preferences as ratings to build predictive models and to generate personalized university ranking lists. In Study 1, we performed offline evaluation on a rating dataset to determine which recommender approaches had the highest predictive value. In Study 2, we selected three algorithms to produce different university recommendation lists in our online tool, asking our users to compare and evaluate them in terms of different metrics (Accuracy, Diversity, Perceived Personalization, Satisfaction, and Novelty). We show that a SVD algorithm scores high on accuracy and perceived personalization, while a KNN algorithm scores better on novelty. We also report findings on preferred university features.
|Service-Oriented Computing - ICSOC 2021 Workshops - AIOps, STRAPS, AI-PA, and Satellite Events, Dubai, United Arab Emirates, November 22-25, 2021, Proceedings. Lecture Notes in Computer Science. Proceeding |
Hakim Hacid; Monther Aldwairi; Mohamed Reda Bouadjenek; Marinella Petrocchi; Noura Faci; Fatma Outay; Amin Beheshti; Lauritz Thamsen; Hai Dong
|Hybrid Recommendation of Movies based on Deep Content Features Inproceedings |
Tord Kvifte; Mehdi Elahi; Christoph Trattner
In: Springer Nature, 2022.
Tags: Cristin| |
When a movie is uploaded to a movie Recommender System (e.g., YouTube), the system can exploit various forms of descriptive features (e.g., tags and genre) in order to generate personalized recommendation for users. However, there are situations where the descriptive features are missing or very limited and the system may fail to include such a movie in the recommendation list. This paper investigates hybrid recommendation based on a novel form of content features, extracted from movies, in order to generate recommendation for users. Such features represent the visual aspects of movies, based on Deep Learning models, and hence, do not require any human annotation when extracted. We have evaluated our proposed technique using a large dataset of movies and shown that automatically extracted visual features can mitigate the cold-start problem by generating recommendation with a superior quality compared to diﬀerent baselines, including recommendation based on human-annotated features.
|The Future Technologies in Journalism Presentation |
EBU Metadata Network 2022 Online Conference, 01.01.2022.
|Deep Learning to Encourage Citizen Involvement in Local Journalism Inproceedings |
Bjørnar Tessem; Lars Nyre; Michel dos Santos Mesquita; Paul Mulholland
In: Palgrave Macmillan, 2022.
|Responsible media technology and AI: challenges and research directions Journal Article |
Christoph Trattner; Dietmar Jannach; Enrico Motta; Irene Costera Meijer; Nicholas Diakopoulos; Mehdi Elahi; Andreas L. Opdahl; Bjørnar Tessem; Njål Borch; Morten Fjeld; Lilja Øvrelid; Koenraad De Smedt; Hallvard Moe
In: AI and Ethics, 2021.
|Towards Responsible Media Recommendation Journal Article |
Mehdi Elahi; Dietmar Jannach; Lars Skjærven; Erik Knudsen; Helle Sjøvaag; Kristian Tolonen; Øyvind Holmstad; Igor Pipkin; Eivind Throndsen; Agnes Stenbom; Eivind Fiskerud; Adrian Oesch; Loek Vredenberg; Christoph Trattner
In: AI and Ethics, 2021.
|Når kunstig intelligens inntar redaksjonen Medium |
|WP3 2021 M3.1 Report The industrial expectations to, needs from and wishes for the work package Technical Report |
Are Tverberg; Ingrid Agasøster; Mads Grønbæck; Marius Monsen; Robert Strand; Kristian Eikeland; Eivind Throndsen; Lars Westvang; Tove B. Knudsen; Eivind Fiskerud; Rune Skår; Sergej Stoppel; Arne Berven; Glenn Skare Pedersen; Paul Macklin; Kenneth Cuomo; Loek Vredenberg; Kristian Tolonen; Andreas L. Opdahl; Bjørnar Tessem; Csaba Veres; Duc-Tien Dang-Nguyen; Enrico Motta; Vinay Jayarama Setty
University of Bergen, MediaFutures 2021.
Publications from 2020 and before are not direct results of the SFI MediaFutures, but are key results from our team members working on related topics in MediaFutures.