The paper’s problem of interest is the identification of an epidemic’s patient zero given a network of contacts and a set of infected individuals, under the assumptions that the infection states of only a few individuals are initially observed and the epidemic has evolved. To tackle this issue, they formulate the problem as an active querying problem and propose a number of active querying strategies inspired by active learning. Their results suggest that in the limited information scenario it is possible to achieve source inference performance comparable to when the infection states of all individuals are observed.

A paper based on controllable models for simplifying medical text, co-authored by our colleague Rosni Kottekulam Vasu and external collaborators, was accepted at the 37th AAAI Conference on Artificial Intelligence! The paper is titled “Med-EASi: Finely Annotated Dataset and Models for Controllable Simplification of Medical Texts” and was jointly conducted with Chandrayee Basu, Michihiro Yasunaga from Stanford University, and Qian Yang from Cornell University.

The paper’s vision is an interactive automatic medical text simplification system, which can enable medical practitioners and patients to simplify the contents of a text or conversation selectively and have controllability over the type of desired textual transformations.

Rosni presented the work at AAAI, orally and on a poster and was accepted to the 2023 AAAI student scholarship and volunteer program. Congratulations!

In his Bachelor Thesis, Viktor Lakic investigated the decay happening in datasets when the resources that Web-URLs point to become unavailable. This Link-Rot can cause problems for reproducibility, as datasets can shrink over time, potentially changing the outcome of experiments which use them. A paper based on the data that Viktor collected in his thesis, co-authored by Luca Rossetto and Abraham Bernstein, was recently presented at the 2023 International Conference on Multimedia Modeling in the Special Session on ‘Multimedia Datasets for Repeatable Experimentation’. The paper was awarded the ‘Best Special Session Paper Award’, honoring the best contribution across all special sessions of the conference. Congratulations!

Current integration approaches are limited to two spaces and presume that both of them were embedded with the same method — both assumptions are unlikely to hold in the context of a Web of Embeddings. In this paper, we present FedCoder— an approach that integrates multiple embedding spaces via a latent space. We assert that linked entities have a similar representation in the latent space so that entities become comparable across embedding spaces. FedCoder employs an autoencoder to learn this latent space from linked as well as non-linked entities.

Our experiments show that FedCoder substantially outperforms state-of-the-art approaches when faced with different embedding models, that it scales better than previous methods in the number of embedding spaces, and that it improves with more graphs being integrated whilst performing comparably with current approaches that assumed joint learning of the embeddings and were, usually, limited to two sources. Our results demonstrate that FedCoder is well adapted to integrate the distributed, diverse, and large ecosystem of embeddings spaces into an interoperable Web of Embeddings.

You can read the full article here.

The paper, which looks into how the kind of expert giving the advice — i.e., a human or an AI advisor — influences trust, perceived responsibility, and reliance.

For more information about the paper, please check it out in our on-line library, over at the ACM Digital Library, and the video presentation prepared by Suzanne.

For your convenience you can also find the abstract right here:

While artificial intelligence (AI) is increasingly applied for decision-making processes, ethical decisions pose challenges for AI applications. Given that humans cannot always agree on the right thing to do, how would ethical decision-making by AI systems be perceived and how would responsibility be ascribed in human-AI collaboration? In this study, we investigate how the expert type (human vs. AI) and level of expert autonomy (adviser vs. decider) influence trust, perceived responsibility, and reliance. We find that participants consider humans to be more morally trustworthy but less capable than their AI equivalent. This shows in participants’ reliance on AI: AI recommendations and decisions are accepted more often than the human expert’s. However, AI team experts are perceived to be less responsible than humans, while programmers and sellers of AI systems are deemed partially responsible instead.

“Evaluation of Algorithms for Interaction-Sparse Recommendations: Neural Networks don’t Always Win“

Authors: Yasamin Klingler, Claude Lehmann, João Pedro Monteiro, Carlo Saladin, Abraham Bernstein, Kurt Stockinger

Abstract: “In recent years, top-K recommender systems with implicit feedback data gained interest in many real world business scenarios. In particular, neural networks have shown promising results on these tasks. However, while traditional recommender systems are built on datasets with frequent user interactions, insurance recommenders often have access to a very limited amount of user interactions, as people only buy a few insurance products. In this paper, we shed new light on the problem of top-K recommendations for interaction-sparse recommender problems.
In particular, we analyze six different recommender algorithms, namely a popularity-based baseline and compare it against two matrix factorization methods (SVD++, ALS), one neural network approach (JCA) and two combinations of neural network and factorization machine approaches (DeepFM, NeuFM). We evaluate these algorithms on six different interaction-sparse datasets and one dataset with a less sparse interaction pattern to elucidate the unique behavior of interaction-sparse datasets.
In our experimental evaluation based on real-world insurance data, we demonstrate that DeepFM shows the best performance followed by JCA and SVD++, which indicates that neural network approaches are the dominant technologies. However, for the remaining five datasets we observe a different pattern. Overall, the matrix factorization method SVD++ is the winner. Surprisingly, the simple popularity-based approach comes out second followed by the neural network approach JCA. In summary, our experimental evaluation for interaction-sparse datasets demonstrates that in general matrix factorization methods outperform neural network approaches. As a consequence, traditional well-established methods should be part of the portfolio of algorithms to solve real-world interaction-sparse recommender problems.”

The paper can be read here.

“HyText – A Scene-Text Extraction Method for Video Retrieval.“

Authors: Alexander Theus, Luca Rossetto, and Abraham Bernstein

Abstract: “Scene-text has been shown to be an effective query target for video retrieval applications in a known-item search context. While much progress has been made in scene-text extraction from individual pictures, the special case of video has so far received less attention. This paper introduces HyText, a scene-text extraction method for video with a focus on retrieval applications. HyText uses intermittent scene-text detection in combination with bi-directional tracking in order to increase throughput without reducing detection accuracy.”

You can read the full paper here.

The first publication is a journal article published in the distinguished Digital Journalism journal. Written as a collaboration between computer science and communication science scholars, the article studies the effects of diverse news recommender systems in the context of democracy.

“Benefits of Diverse News Recommendations for Democracy: A User Study“

Authors: Lucien Heitz, Juliane A. Lischka, Alena Birrer, Bibek Paudel (former DDIS member, currently at Stanford University), Suzanne Tolmeijer (former DDIS member, soon at the University of Hamburg), Laura Laugwitz, and Abraham Bernstein

Abstract: “News recommender systems provide a technological architecture that helps shaping public discourse. Following a normative approach to news recommender system design, we test utility and external effects of a diversity-aware news recommender algorithm. In an experimental study using a custom-built news app, we show that diversity-optimized recommendations (1) perform similar to methods optimizing for user preferences regarding user utility, (2) that diverse news recommendations are related to a higher tolerance for opposing views, especially for politically conservative users, and (3) that diverse news recommender systems may nudge users towards preferring news with differing or even opposing views. We conclude that diverse news recommendations can have a depolarizing capacity for democratic societies.”

You can read the entire article here.

The second publication is the book “Spotlight on Artificial Intelligence and Freedom of Expression: A Policy Manual“, co-authored by Eliska Pirkova, Matthias Kettemann, Marlena Wisniak, Martin Scheinin, Emmi Bevensee, Katie Pentney, Lorna Woods, Lucien Heitz, Bojana Kostic, Krisztina Rozgonyi, Holli Sargeant, Julia Haas, and Vladan Joler. You can read the full book here.

While this great achievement provides cause for celebration, we are also sad to see Romana leave to start her position as a postdoctoral researcher at the Vrije Universiteit Amsterdam in the Netherlands as soon as September 1! We wish her good luck and all the best for her future endeavors!

At the same conference, a group of DDIS members also presented a demo of a graph-based retrieval system for lifelogs. ‘LifeGraph’ was developed in the context of an international evaluation campaign which, due to some changes and delays caused by the pandemic, was held online only a week before ISWC. During the demo, conference attendees could get an impression of how such a graph exploration-based approach could be used to find particular events from the life of a lifelogger. The demo was nominated for ‘Best Demo’. More information on the inner workings of that approach can be found in the paper, authored by Luca Rossetto, Matthias Baumgartner, Narges Ashena, Florian Ruosch, Romana Pernisch, and Abraham Bernstein.