UMAP '23 Adjunct: Adjunct Proceedings of the 31st ACM Conference on User Modeling, Adaptation and Personalization

Full Citation in the ACM Digital Library

SESSION: Late-Breaking Results and Demos

A Multi-factorial Analysis of Polarization on Social Media

Polarization is an increasingly worrying phenomenon within social media. Recent work has made it possible to detect and even quantify polarization. Nevertheless, the few existing metrics, although defined in a continuous space, often lead to a unimodal distribution of data once applied to users’ interactions, making the distinction between polarized and non-polarized users difficult to draw. Furthermore, each metric relies on a single factor and does not reflect the overall user behavior. Modeling polarization in a single form runs the risk of obscuring inter-individual differences. In this paper, we propose to have a deeper look at polarized online behaviors and to compare individual metrics. We collected about 300K retweets from 1K French users between January and July 2022 on Twitter. Each retweet is related to the highly controversial vaccine debate. Results show that a multi-factorial analysis leads to the identification of distinct and potentially explainable behavioral classes. This finer understanding of behaviors is an essential step to adapt news recommendation strategies so that no user gets locked into an echo chamber or filter bubble.

Addressing Popularity Bias in Recommender Systems: An Exploration of Self-Supervised Learning Models

The rapid growth of the volume and variety of online media content has made it increasingly challenging for users to discover fresh content that meets their particular needs and tastes. Recommender Systems are digital tools that support users in navigating the plethora of available items. While these systems may offer several benefits, they may also create or reinforce certain undesired effects, including Popularity Bias, i.e., a short list of popular items becoming more popular while a long list of unpopular ones becoming even more unpopular.

In this paper, we focus on this challenge and propose a novel recommendation approach that can generate accurate recommendations while effectively mitigating the popularity bias. Our proposed approach adopts models based on Self-Supervised Learning (SSL) that have recently drawn considerable attention in various application domains. Such models are known to enable recommender systems to exploit automatic mechanisms for data annotation hence providing self-supervisory signals for better training of the system from the available data. We considered various recommendation techniques based on the SSL model and compared their impact on popularity bias mitigation measured in terms of Average Recommendation Popularity (ARP), Gini-index, and Coverage. The results showed that SSL models could successfully mitigate the popularity bias while still maintaining the accuracy of the recommendation.

An Out-of-the-Box Application for Reproducible Graph Collaborative Filtering extending the Elliot Framework

Graph convolutional networks (GCNs) are taking over collaborative filtering-based recommendation. Their message-passing schema effectively distills the collaborative signal throughout the user-item graph by propagating informative content from neighbor to ego nodes. In this demonstration, we show how to run complete experimental pipelines with six state-of-the-art graph recommendation models in Elliot (i.e., our framework for recommender system evaluation). We seek to highlight four main features, namely: (i) we support reproducibility in PyTorch Geometric (i.e., the library we use to implement the baselines); (ii) reproduced graph models span across various GCN families; (iii) we prepare a Docker image to provide a self-consistent ecosystem for the running of experiments. Codes, datasets, and a video tutorial to install and launch the application are accessible at:

Chai Wallpaper: A Mindfulness-Based Persuasive Intervention for Absent-Minded Smartphone Usage

Digital mindfulness applications and tools are increasingly developed to improve the wellbeing of technology users. In this study, we present Chai, an Android live wallpaper application designed to reduce absent-minded smartphone usage. The wallpaper is designed as a tree whose leaves undergo color change and fall off based on the frequency of smartphone use. We also present the result of an initial pilot study conducted using the application. The participants reported increased awareness of their smartphone use behavior and found the wallpaper and the other features of the application to be persuasive at varying degrees.

ChatGPT in the Classroom: A Preliminary Exploration on the Feasibility of Adapting ChatGPT to Support Children’s Information Discovery

The influence of ChatGPT and similar models on education is being increasingly discussed. With the current level of enthusiasm among users, ChatGPT is envisioned as having great potential. As generative models are unpredictable in terms of producing biased, harmful, and unsafe content, we argue that they should be comprehensively tested for more vulnerable groups, such as children, to understand what role they can play and what training and supervision are necessary. Here, we present the results of a preliminary exploration aiming to understand whether ChatGPT can adapt to support children in completing information discovery tasks in the education context. We analyze ChatGPT responses to search prompts related to the 4th grade classroom curriculum using a variety of lenses (e.g., readability and language) to identify open challenges and limitations that must be addressed by interdisciplinary communities.

Covering Covers: Characterization Of Visual Elements Regarding Sleeves

The aim of this work is to explore common traits preferred across different age groups of children to identify the appeal of book covers. By analyzing visual attributes, visible objects, and implied stories inferred from the covers, we can gain insights into the elements that are most attractive to children up to 18 years old. These findings can then contribute towards advancing personalization for recommender systems for children through new means that do not rely on historical data, seldom available for this user group.

Cyclops: Looking beyond the single perspective in information access systems

Information access systems — such as search engines and social media — filter, sort, and otherwise control the content that people are exposed to. Uncritical consumption of information online can lead to extremism, cognitive bias, etc., so it is crucial for people to question the information with which they are presented. It is even more important in the case of young people, as they start learning about the world and form their own opinions. Therefore it is important that people are made aware of how information access systems function, and the related potential risks of uncritical consumption of information. We propose an interactive tool to encourage critical consumption of online information, that acts also as a research tool on the experience and knowledge of people in relation to information access systems.

Development of a Trust-Aware User Simulator for Statistical Proactive Dialog Modeling in Human-AI Teams

The concept of a Human-AI team has gained increasing attention in recent years. For effective collaboration between humans and AI teammates, proactivity is crucial for close coordination and effective communication. However, the design of adequate proactivity for AI-based systems to support humans is still an open question and a challenging topic. In this paper, we present the development of a corpus-based user simulator for training and testing proactive dialog policies. The simulator incorporates informed knowledge about proactive dialog and its effect on user trust and simulates user behavior and personal information, including socio-demographic features and personality traits. Two different simulation approaches were compared, and a task-step-based approach yielded better overall results due to enhanced modeling of sequential dependencies. This research presents a promising avenue for exploring and evaluating appropriate proactive strategies in a dialog game setting for improving Human-AI teams.

Eudaimonic and Hedonic Qualities as Predictors of Music Videos’ Relevance to Users: A Human-Centric Study

In this study, we investigated if the user’s eudaimonic and hedonic orientation (EHO) and the eudaimonic and hedonic (EH) scores of a music video contribute to a successful prediction of user preferences. The study was carried out on real content and users in collaboration with the Dutch music video streaming platform XITE. We collected EH annotations from the music expert curator team at XITE and conducted a user study to collect users’ data, including EHO. Using machine learning models, we predicted the relevance of a music video to a user based on the EHO and EH scores. Our results confirmed the hypothesized relationship, with several models outperforming the baseline.

Expected and Experienced Utility of Points of Interest in Tourism Recommender Systems

In this paper we analyze users’ preferences for Points of Interest (POIs), before and after the POIs are experienced: these are called expected and experienced utilities. State-of-the-art research in other domains, has shown that the difference between the user’s expected utility, which is estimated at the choice stage, and the experienced utility, obtained after the consumption of the item, is domain dependent. This discrepancy is here measured in the tourism domain by focusing on the POI recommendation problem. We have designed and implemented an interactive web-based survey to collect the two types of utility data. The analysis of the data collected from a sample of users shows that the two utilities express quite different preferences. This result may be exploited to design novel POI recommendation models that can optimally leverage both types of preference signals.

Extending Bayesian Personalized Ranking with Survival Analysis for MOOC Recommendation

Massive Open Online Courses (MOOCs) have recently attracted students and professionals as complementary tools to academic education. Despite the number of advantages MOOCs provide, such as openness and flexibility regarding learning pace, such courses are characterized by a consistently higher dropout rate than conventional classrooms. A crucial factor that influences dropout is the choice of the appropriate course, hence the need for effective course recommendations. A course recommendation system (RS) that uses dropout information can mitigate course withdrawal and user dissatisfaction. In this paper, an extension of Bayesian Personalized Ranking, which is a learning-to-rank RS, is proposed that uses the pseudo-labels extracted by survival analysis based on dropout information to recommend courses in the context of MOOCs. The proposed approach performs the best compared to six competing RSs on three MOOCs datasets.

How can we model climbers’ future visits from their past records?

Outdoor sport climbing is one of the major outdoor sports in Northern Italy due to the vast number of rock climbing places named crags. New crags appear yearly, creating an information overload problem for climbers when they plan where to climb on their forthcoming trips. As such, climbing crags recommender systems address this problem, suggesting crags based on the number of routes similar to those liked by the user in the past. At the same time, people are interested not only in the climbing routes but they have other objectives; for instance, they plan to teach their children to climb while traveling, or prefer to visit the place only if it has a parking space since they travel by car. To better understand and model future climbers’ choices in outdoor sport climbing environments, we first define crags characteristics that primarily impact a user’s behavior. Secondly, we propose to model climbers’ profiles and their preferences for crags’ characteristics as Pearson correlation computed with the number of past users’ visits. Thirdly, we developed and evaluated the recommender system called Visit&Climb, where user tastes are projected into an interactive preferences elicitation panel, which users can further employ to adjust their profile with the sliders. The recommendations are then supplied as the top-visited crags by the most similar user. For the evaluation, we ran several offline experiments: we compared different models (regression-based, matrix factorization, and collaborative filtering) to predict visits recorded in 99 crags by 106 climbers in Arco (Italy). During these experiments, we measured standard metrics such as MaP@k, Recall@k, and NDCG@k for top-k ranking quality. The offline evaluation showed that the Visit&Climb system provides more accurate recommendations than the Baseline model, which utilizes users’ previous records for future prediction. Plus, it delivers a comparable accuracy level to other systems in this domain. Moreover, unlike the other solutions, this developed method visualizes the users’ profiles and allows modification of their tastes, solving a cold-start issue. The recommender system proposed in this work can confidently model climbers’ future visits by their past logs.

Image-based Face Verification for Student Identity Management — the TRUSTID Case Study

Managing attendance and confirming student identity in online lessons is a contemporary problem of Higher Education Institutions (HEIs) that comes with its challenges, especially in classes with large numbers of students. The development and deployment of fully automatic and continuous image-based face verification techniques may provide a natural solution for this problem and assist instructors in handling the complexity of managing student identity in a remote classroom.

In this paper, we present an automatic image-based student identification framework. Our approach proposes a biometric authentication platform using a Residual Network (ResNet) Learning method for face verification. The system described is intended for use with consumer grade web cameras, providing a tradeoff between reliability and computational performance, as no assumption can be done regarding the target student hardware (i.e. CPU or GPU). This follows from the case study of TRUSTID, an European R&D initiative for intelligent student identity management in distance learning scenarios.

MisinfoMe: A Tool for Longitudinal Assessment of Twitter Accounts’ Sharing of Misinformation

Persistent and widespread misinformation continues to pose a threat to societies on various levels. Despite the concerted efforts to address this issue, the challenge of capturing and scrutinising the interaction of individuals with misinformation remains. In this paper, we introduce MisinfoMe; a tool that leverages Fact-Checkers’ ClaimReview annotations and source-level validations to assess the credibility of Twitter accounts based on their sharing of misinformation over time.

Modelling Student Knowledge in Blended Learning

Blended learning offers a diverse learning experience through multiple activities inside and outside the classroom, which can improve student knowledge, as there are multiple opportunities for learning. However, managing these activities requires an integrated approach to ensure its effectiveness, that is, taking into account learning data from different sources. Disregarding any of these sources may lead to incomplete/incorrect information on the current levels of students’ understanding of courses topics. This paper proposes an approach to student modelling that incorporates both streams of student activity performed during both modes of blended learning. To maintain a mode meaningful representation of students’ knowledge, reflecting differences in focuses of in-class and at-home assessment, the proposed approach divides student knowledge into three cognitive levels based on Bloom’s taxonomy, namely, Remember, Understand, and Apply. The Elo Rating System is used as the main method of student knowledge estimation; it is enriched with knowledge propagation between the Bloom’s levels of cognitive activity to account for their inter-dependency. The propagation parameters are optimised. The result shows that the model is capable to distinguish between positive and negative results of student attempts well enough.

Multi-Criteria Ranking by Using Relaxed Pareto Ranking Methods

Multi-criteria recommender systems can improve the quality of recommendations by considering user preferences on multiple criteria. One promising approach proposed recently is multi-criteria ranking, which uses Pareto ranking to assign a ranking score based on the dominance relationship between predicted ratings across criteria. However, applying Pareto ranking to all criteria may result in non-differentiable ranking scores. To alleviate this issue, we conducted a study on three relaxed Pareto ranking methods for multi-criteria ranking. We evaluated these methods on three real-world datasets and found that the k-dominance ranking approach, which is one of the relaxed Pareto ranking methods, was able to further enhance the ranking performance.

Personalized Learning Systems for Computer Science Students: Analyzing and Predicting Learning Behaviors Using Programming Error Data

The integration of technology in education has become indispensable in acquiring new skills, knowledge, and competencies. This paper addresses the issue of analyzing and predicting the learning behavior of Computer Science students. Specifically, we present a dataset of compiler errors made by students during the first semester of an Introduction to Programming course where they learn the C programming language. We approach the problem of predicting the number of student errors as a missing data imputation problem, utilizing several prediction methods including Singular Value Decomposition, Polynomial Regression via Latent Tensor Reconstruction, Neural Network-based method, and Gradient Boosting. Our experimental results demonstrate high accuracy in predicting student learning behaviors over time, which can be leveraged to enhance personalized learning for individual students.

SpotifyExplained: User-centric Mobile Application for Music Exploration

Many music streaming services, such as Spotify, contain a large volume of heavy users, whose user profiles may contain tens to hundreds artists or hundreds to thousands favorite tracks. For those users, it may be difficult to understand, what data the service collected about them, what is the base for its automated decisions (e.g., recommendations) and why are particular items suggested to them. This may decrease user’s willingness to explore the suggested content and may eventually lead to the decrease of user satisfaction and trust. In this demo, we present SpotifyExplained Android application, which is built on top of Spotify API and aims to bridge the aforementioned understandability gap. Specifically, the application contains several tunable views on users’ music profiles and visually explain the connections between both the already known content as well as newly suggested items.

Supporting a Group Member to Make a Group Choice

We often make choices that involve a group of people, such as selecting a movie to watch with friends or choosing a travel destination to visit with the family. Sometimes, a single member of the group may be in charge of making the decision for the group, by playing the role of “organizer”. Although some tools for supporting Group Decision-Making have been proposed, none of them have considered the case where a single group member is autonomously making such a decision, hence entering the preferences of the group members, interacting with the system, and finally selecting a proper recommendation. In this paper, we introduce MyFoodGRS, a web application for a single user to find a proper restaurant for their group, that supports the previously mentioned tasks. We introduce an interaction design to follow the Attribute and Socially-based group decision patterns, and we report the positive result of the conducted system usability evaluation.

The Impact of Personalised Advertisement Campaigns on Tourist Choices in South Tyrol: A Sustainable Tourism Perspective

Overtourism, i.e., the excessive presence of tourists in a location, is a widespread problem. To tame it, it is essential to understand tourism trends (tourists arrivals in the region areas) but also to predict the effects that personalised marketing campaigns, which are routinely performed by destination management organisations, may have on such trends. To facilitate this analysis, we propose a simulation system and we showcase its application to South Tyrol, a tourism region in the Italian Alps. The performed simulations are based on actual and feature-rich tourism arrivals data, collected in the past ten years. We simulate that the logged tourists are exposed to some advertised districts and their consequent choices. The demoed system enables the analysis of tourism data, the set up of marketing effect simulations, and the visualisation of their results. The system may reveal the broad effect of personalised and non-personalised advertisements, hence can help to identify which marketing campaigns are more likely to improve tourism sustainability.

Trust-based Recommender System for Fake News Mitigation

The ubiquity of fake news has been a serious problem on the Internet. Recommender systems, in particular, contribute to this issue by creating echo chambers of misinformation. In light of these observations, we address the issue of fake news mitigation through the lens of recommender systems. This paper introduces a novel adaptation of the collaborative filtering algorithm that models untrustworthy online users in order to remove them from the candidate user’s neighborhood. The proposed approach, FAke News Aware Recommender system (FANAR), is an alteration of the collaborative filtering strategy that considerably prevents the propagation of fake news by avoiding untrustworthy neighbors. Furthermore, we create FNEWR, a dataset for the Fake News Recommendation system, to fulfill our goal. Our experiments reveal that FANAR surpasses the current leading news recommendation techniques in its ability to suggest personalized news and mitigate the spread of false information.

TRUSTID: Intelligent and Continuous Online Student Identity Management in Higher Education

Online learning and remote assessment of students within Higher Education Institutions (HEIs) have shown significant growth over the last few years since the COVID-19 outbreak. In this context, the shift of various HEIs to online teaching and evaluation introduced several challenges, especially with relation to student identification and interaction behavior within critical online academic activities (e.g., examinations). To overcome these challenges, existing solutions relied on a combination of an online examination through a learning management system (LMS) with a basic user authentication method for identification of students, and video conference tools for manual monitoring of students. However, such solutions fall short in dealing with various threat scenarios, and cannot be easily integrated and modified by HEIs. In this paper, we present the development and initial usability and user experience evaluation of the proof of concept of a multi-tier continuous user identification framework, bootstrapped to HEI contexts, that consists of state-of-the-art intelligent image and voice biometrics. The suggested solution is currently being implemented and evaluated by the European Commission as part of the actions of ERASMUS+ 2020 and in particular the Call “Strategic Partnerships in Response to the COVID-19 Situation: Partnerships for Digital Education Readiness in the field of Higher Education (KA226)”.

United-and-Close: An interactive visual platform for assessing urban segregation within the 15-minutes paradigm

The ‘15-minute city’ paradigm is an urban model based on the concept of ‘hyper-proximity’: citizens should be able to access fundamental services and facilities (such as schools, shops, parks, doctors, and markets) within 15-20 minutes on foot, by bicycle or by public transport. Compliance with the ‘15-minute city’ paradigm is supposed to reduce pollution and social inequalities. It is supposed to bring the psychological fragility of the citizen back to the center of the urban redevelopment debate. Although the concept has gained great attention and interest from policymakers and urban designers, we still lack tools that can help to validate, on a data-driven basis, the assumption that hyper-proximity is eventually correlated with lower urban segregation, which is one of the driving forces that lead to social inequalities. We aim to define a data-driven methodology to analyze the urban areas where services should be accessible within 15 minutes; network analysis is exploited to estimate services proximity as well as the connectivity of different urban areas with each other, in order to gather signals of the general resilience or exposure to urban segregation. We also aim to compute a set of city-agnostic metrics that will include user-specified parameters and personalized weights for each Point of Interest’s category. United-and-Close is the resulting Web platform designed to be accessible to citizens, policy and decision-makers, and investors, but also for researchers involved in disciplines such as urban informatics that need support to better assess the 15-minute paradigm and its actual impact on our cities.

SESSION: ADAPPT 2023: Adaptive and Personalized Persuasive Technologies Workshop

Adaptive and Personalized Persuasive Technologies (ADAPPT 2023) Workshop

The Adaptive and Personalized Persuasive Technologies (ADAPPT’23) workshop holding in Cyprus this year is the third edition of the ADAPPT series, which commenced in 2019. The workshop is organized in conjunction with the 31st Association for Computer Machinery (ACM) Conference on User Modeling, Adaptation and Personalization (UMAP). In this preface to the third edition, we summarize the papers accepted for publication in the adjunct proceedings. Finally, we present a list of the members of the organizing and program committees that made the workshop a success.

A Usability Evaluation of a Software Framework for Designing Persuasive Games.

With the rise of persuasive game design for health, there is a need for easy, quick, and effective ways of developing and testing out persuasive games across multiple domains. This paper presents the design and usability evaluation of P-Gamer – a software framework for developing persuasive games and evaluating the effectiveness of various strategies. In line with the user-centred design approach, we designed a prototype of the system and conducted a usability evaluation with six persuasive system designers to (a) understand how usable the P-Gamer platform is, (b) understand the ease of use for each of the six major features in the platform, and (c) identify and correct the design issues existing in the platform. Our results showed that the overall system was perceived to be useable. The system had an overall system usability score of 87.92, which is within the excellent score range in the SUS scale. Our results also showed that five of the six major sections of the platform were significantly easy to use. We reflect on the results and also discuss the design issues and insights into addressing them.

Adaptive and Privacy-Aware Persuasive Strategies to Promote Healthy Eating Habits: Position Paper

Food has become a social issue. Developing a healthy relationship with food is a complex process as it implies a behaviour change. At first sight, recent advances in recommender system domain ensure personalised service and assistance of a user in decision-making process. However, traditional recommender systems lack the consideration of health-related factors (e.g., healthy diet guidelines, food allergies, etc.). Moreover, instead of recommending a new item (recipe), an effective system should apply persuasion strategies to accompany a user in a long process, a little step at a time. Furthermore, most of existing recommender models do not take privacy preservation into account. In this position paper, we outline the challenges faced by a persuasive system for promoting eating habits and discuss our vision on the implementation of adaptive and privacy-aware persuasive strategies for healthy food promotion.

Context-Dependent Use of Authority and Empathy in Lifestyle Advices Given By Persuasive Voice Assistants

As smart technology becomes more readily available for the general public, so do the systems that can benefit of such technology, such as recommender systems and Voice Assistants. Persuasive Voice Assistants have a great potential to change people’s lives by promoting changes in behavior that benefit people’s health and lifestyles. Previous studies have focused on finding personality traits that systems can use to optimize persuasive capabilities, but the findings have been contradictory. In this study, we investigate these contradictions by comparing the effect of empathic phrasing versus authoritative phrasing of persuasive recommendations in three lifestyle-related domains. The study results as well as the literature survey strongly indicate that it is close to impossible to isolate general effects, emphasizing the importance of taking all specifics of the user population, individual differences and user context as whole – as well as interactions between these variables – into account when designing persuasive adaptive systems.

Design framework for the development of tailored behavior change technologies

Today, so-called persuasive or motivational technologies are developing, which refer to technologies, applications or services designed to induce changes in attitudes and behavior in those who use them. This is the subject of research in Human-Computer Interaction in connection with theories from psychology related, for example, to behavior change or motivation. Research on these technologies suggests that in order to encourage long-term adherence, these "virtual coaches" need to be personalized and/or tailored according to the individual characteristics of the users (stage of behavior change, motivations, preferences, barriers). For example, according to the Self-Determination Theory (SDT), an individual can present different forms of motivation, more or less effective. The idea is to be able to identify the forms of motivation present in users in order to propose services to reinforce or develop them. The use of such systems in the field of health has the potential to induce and reinforce health behaviors that are sometimes difficult to establish by health professionals. Providing daily, personalized care represents a considerable human and financial cost for health professionals. A mobile application has the advantage of being able to deal with these constraints. The computer coach is designed with and for patients suffering from chronic low back pain. The aim is to help them manage their condition, particularly with regard to their pain and the practice of regular physical activity. The work presented presents an innovative design approach, combining user-centered methodologies and psychological theories, which we detail through a research phase and a design phase. Currently, a first version of the application has been developed and is being tested.

Factors Influencing the Willingness to Download Contact Tracing Apps among the American Population

The COVID-19 pandemic, which began in 2020, necessitated the roll-out of contact tracing apps (CTAs) worldwide. However, there is limited work to understand the various factors that influence the adoption of CTAs, and the moderating effect of persuasive design and smartphone usage experience. Consequently, we carried out an empirical study among the American population (n = 160) using screenshots of key interfaces of an existing CTA available on the market. The interfaces comprise two types of designs: persuasive version (equipped with persuasive features) and control version (unequipped with persuasive features). Our Partial Least Square Path Model showed that perceived usefulness (β = 0.23, p < 0.05) and perceived compatibility (β = −0.33, p < 0.01) are important to downloading the control version of the app. However, perceived trust (β = 0.45, p < 0.05) is important to downloading the persuasive version of the app. Moreover, perceived usefulness (β = 0.26, p < 0.05) and perceived risk (β = −0.26, p < 0.05) is important for the high-experience group (people with more than 10 years of experience in smartphone usage) and perceived trust (β = 0.46, p < 0.05) for the low-experience group. We discuss the implications of the findings.

Susceptibility of Online Users to Persuasive Strategies to Curb the Spread of Misinformation

The spread of misinformation in online social media is a cause of concern to stakeholders. With many people going online for information about many aspects of their lives, any misinformation presented online can have detrimental effects on those that consume it. The use of persuasive strategies to curb misinformation online is an ongoing research area. There is little or no knowledge of how persuasive strategies can be applied to the different types of social media users to curb the spread of misinformation. To contribute to research in this area, we present the preliminary results of an ongoing user study of currently 113 social media users which investigates which type of social media users will likely spread misinformation and what persuasive strategies will likely influence the different types of social media users. We developed and tested a global model using structural equation modelling. Our results suggest that people that use social media because of friendship and role-playing are influenced by social proof while people who use social media to seek for relationships are influenced by liking and are likely to spread misinformation in the future.

The Influence of Culture in Contact Tracing App Design: A Comparative Analysis of Canada’s COVID Alert vs. India’s Aarogya Setu

The 2020 pandemic culminated in the roll-out of contacting tracing apps in many countries to curb the spread of COVID-19. In this paper, we investigated how culture might have influenced the design of these apps by comparing two national apps (COVID Alert and Aarogya Setu) from two different types of culture: individualist (Canada) and collectivist (India), respectively. We found that there was a cultural effect on the design of both apps, starting from their naming. While the Canada’s app was explicitly named “COVID Alert," the Indian app was implicitly named “Aarogya Setu" (meaning "the bridge to health"). Other key attributes on which both apps differ include number of services, information density, privacy design, use of color, use of metaphors, and multimodality. We discuss these differences and the possible rationale behind each app’s design choices using four key cultural dimensions from Hofstede’s and Hall’s cultural frameworks as analytical lenses. The main contribution of the paper is that it is the first to demonstrate the manifestation of national cultures in the interface and functional design of contact tracing apps by focusing on a Western country and an Asian country.

SESSION: CRUM 2023: The First Edition of the Workshop on Context Representation in User Modelling

1st Workshop on Context Representation in User Modelling

Context, a critical aspect of personalisation and human-computer interaction, is becoming increasingly significant with the proliferation of user-centric applications. The First ACM Workshop on Context Representation in User Modelling (CRUM 2023), organised in conjunction with the 31st ACM Conference on User Modeling, Adaptation and Personalization (UMAP 2023), provided a venue for research in the representation and utilisation of contextual information in intelligent systems. The workshop aimed to bring together researchers from various disciplines, including natural language processing, user modelling, explainable systems, and human-robot interaction and to facilitate conversation about the role of context in adaptive applications. The prospective authors were invited to submit papers of up to seven pages, four of which were accepted for publication and are introduced in this summary.

A Neural Bag-of-Words Point Process Model for User Return Time Prediction in E-commerce

Monitoring and predicting user engagement is essential to gauge the overall health of an E-commerce platform. A healthy active user base indicates that the platform is able to retain users and is performing well on the user satisfaction metric. To measure long-term user satisfaction, predicting the return rate of a user is essential. The frequent return of the user indicates that they are overall satisfied with the platform. To this end, we consider the problem of predicting users’ return time on the platform given their historical interactions.

The current state-of-the-art models for user return time prediction are based on recurrent neural network, which models the sequence of user interactions and predicts the return time using a Temporal Point Process based formulation. However, it is well-known that the inference time for these models grows as the sequence length increases, due to the complex recurrent and gating mechanisms, which deems them unfit to be used in a real-time prediction setting. Towards this end, we propose a lightweight and simple neural bag-of-words-based model for user return time prediction, which considers the user activity trail as a bag-of-words embedding model and performs a simple aggregation operation, used for the final prediction. We perform experiments on interaction log data from a major e-commerce company, and our proposed bag-of-words model outperforms the complex recurrence-based neural network by 6.14% and 4.81% on average, in terms of the Root-Mean-Squared-Error and Mean-Absolute-Error, respectively. We also compare the inference time of our method to the recurrent neural network-based method, with an overall reduction of 78.5% in terms of the wall-clock time.

Characterizing the Emotional Context Induced by Music Listening and its Effects on Gait Initiation: Exploiting Physiological and Biomechanical Data

The ability to initiate gait involves a complex coordination between posture and movement, known as anticipatory postural adjustments (APAs). The emotional context in which gait initiation occurs can impact several spatio-temporal parameters, particularly the duration of APAs. While previous studies have used biologically relevant stimuli to induce emotions, such as images of pleasant or unpleasant scenes, to the best of our knowledge, the impact of the emotional context induced by music on gait initiation has not been explored yet. This paper presents a new dataset collected to study this impact. Objective biomechanical and physiological data were collected from participants during and after music listening, and subjective emotional responses were assessed using questionnaires. We also focused on two factors, liking judgment and familiarity, known to modulate emotions. Our preliminary analyses shows the impact of the emotional context induced by music on gait initiation, and confirms the strong importance of liking judgment and familiarity on the emotional context.

Let’s Talk About The Experienced Context: An Example Regarding Public Transport Information Systems

This position paper aims to encourage researchers in the field of context-aware public transport information systems to incorporate human-centred approaches more deeply into their methodologies. Current context-aware systems in this domain often take a representational view and employ a data-first approach. Drawing on insights from previous work, we propose a distinction between the objective context and the experienced context. The experienced context incorporates interactions and perceptions to reflect better how we, as humans, experience the world. To measure this experienced context, we advocate for using qualitative research methods for HCI. To demonstrate this approach, we present the results of a focus group study on context in public transport. The results reveal that emerging experiences are shaped by a combination of various factors. These findings highlight the importance of incorporating user perspectives in designing context-aware systems. Therefore, in this paper, we take the position that if we want to improve the context-aware public transport information systems, we need to understand what travellers truly experience during their journey.

Make your next item recommendation model time sensitive

Recommender systems have become increasingly popular for providing personalized recommendations to users. Recent studies have shown that transformer-based approaches can enhance the performance of these systems. However, these models usually consider the sequence of past user interactions and do not take into account the time of prediction. In this paper, we address this issue by proposing a simple yet effective method in the form of adapter to make next-item recommenders time-aware. Specifically, we introduce a novel approach that incorporates time information into the modeling process. We conduct extensive experiments on two commonly used sequential recommenders, GRU4Rec and TiSASRec, using four real-world datasets. Our results demonstrate that our approach increases the quality of existing methods and improves the accuracy of recommendations. Our approach is easy to implement and can be applied to a wide range of next-item recommendation systems. It provides a structured framework for incorporating time information into the modeling process, making it easier for researchers to replicate and build upon our findings. Overall, our work contributes to the development of more accurate and efficient recommendation systems, with potential applications in various domains such as e-commerce, social media, and online content delivery. Code is available at GitHub repo1.

SESSION: ExUM 2023: 5th Workshop on Explainable User Models and Personalised Systems

5th Workshop on Explainable User Models and Personalised Systems (ExUM)

Adaptive and personalized systems have become pervasive technologies, gradually playing an increasingly important role in our daily lives. Indeed, we are now used to interacting with algorithms that help us in several scenarios, ranging from services that suggest music or movies to personal assistants who proactively support us in complex decision-making tasks. As the importance of such technologies in our everyday lives grows, it is fundamental that the internal mechanisms that guide these algorithms are as clear as possible. It is not by chance that the EU General Data Protection Regulation (GDPR) emphasized the users’ right to explanation when people face intelligent systems. Unfortunately, current research tends to go in the opposite direction since most of the approaches try to maximize the effectiveness of the personalization strategy (e.g., recommendation accuracy) at the expense of model explainability. The workshop aims to provide a forum for discussing problems, challenges, and innovative research approaches in this area by investigating the role of transparency and explainability in recent methodologies for building user models or developing personalized and adaptive systems.

Branching Preferences: Visualizing Non-linear Topic Progression in Conversational Recommender Systems

Recent advances in AI allow complex, natural user–system dialogue flow in NLP-based conversational recommender systems (CRS). While this enables users to express complex intents to the system, its usual linear GUI representation as a chat log fails to account for two non-linear aspects of natural conversation: humans can switch between topics as customary; and, especially in decision-making contexts, topics discussed are structurally related. As early work, we motivate and present a GUI design approach that aims to exploit these phenomena for CRS by conveying topic progression, and discuss several design variants, their trade-offs, and open questions. Our approach aims to help users orientate while exploring and comparing multiple preference model variants and corresponding recommendations in complex, natural ways, also accounting for different explanation types. Such orientation could benefit users for achieving complex goals using CRS, like thoroughly-informed decision making, getting inspiration for novel consumable items, and exploring their own preferences.

Computer Vision, Human Likeness, and Problematic Behaviors: Distinguishing Stereotypes from Social Norms

Computer Vision (CV) has become an essential tool for developers looking to personalize user experiences. In particular, commercial CV services can be used by those who are not machine learning experts, but who want to enhance their apps and services with vision capabilities. While the performance of CV has become increasingly human-like, its “social behaviors" and their compatibility with human values are of concern. In contrast to algorithmic decision-making, where fairness is used to evaluate system behavior, CV is often evaluated for stereotyping – the extent to which systems reflect prevalent social beliefs. This paper proposes that viewing stereotyping negatively is unhelpful in improving human-AI interaction. Rather, it is more fruitful to separate the observation of a social behavior (i.e., documenting what a machine does in relation to a human) from its judgment (i.e., relating the behavior to social norms). As norms differ across contexts and application areas, such an approach better reflects the real world, which is characterized by diversity and opposing views. However, it requires us to face up to two truths: i) humans – not machines – are the problem; ii) we must decide what degree of human-likeness we ultimately want; technologies designed to mimic us will reflect social bias.

Ethical issues in explanations of personalized recommender systems

Explanations provide means to increase transparency and trust in personalized recommender systems. However, the design of explanations can raise new ethical problems. This article aims to raise awareness about the ethical considerations that should be taken into account when designing, implementing, and evaluating explanations for personalized recommender systems.

Exploring cognitive models to augment explainability in Deep Knowledge Tracing

Adaptive learning systems allow a personalized adaptation based on the characteristics of the student. Tracing the progress of knowledge and skills during the learning process through cognitive models is essential so that these systems can make appropriate decisions when carrying out personalization. This is the objective of Knowledge Tracing, which studies how to infer a cognitive model from the answers given to a sequence of questions or exercises. The incorporation of Deep Learning techniques in this field has given rise to Deep Knowledge Tracing (DKT) which usually has excellent predictive outcomes. The problem is that this increase in accuracy comes with a lack of explainability since Deep Learning models can be considered black boxes from which it is difficult to build interpretations or explanations. By contrast, traditional Knowledge Tracing methods are based on underlying learning models and provide a solid basis for explainability. In this paper we describe an ongoing research to build DKT models with a good trade-off between accuracy and explainability. To this end, we propose to use a loss function based on a mixup approach where the ground truth is a mix between the dataset labels and the predictions of a surrogate explainable model. The approach has potential to improve, not only explainability through the use of the surrogate, but also accuracy thanks to regularization effects. We will validate the approach by exploring, for different cognitive models, the trade-off curve that is obtained by plotting accuracy against explainability for different mixup values.

Synchronized Multi-list User Interfaces for Fashion Catalogs

Several online catalogs use carousels to present thematic lists of products, based on different optimization criteria. While this makes it possible to search for items according to diverse relevance perspectives, it hardly supports an integrated evaluation, which is key to critical consuming behavior. To address this issue, we propose a synchronized multi-list model that (i) enriches item presentation by visualizing its evaluation and (ii) enables the user to simultaneously center the carousels of the multi-list on the item in her/his focus of attention, showing its ranking in each list. This type of visualization is aimed at enhancing the transparency of results by enabling the user to simultaneously compare products across all the evaluation criteria applied within the multi-list.

As a testbed for our model, we selected fashion catalogs, with the aim of making users aware of clothes’ evaluation with respect to the sustainability and ethical issues concerning the production practices applied by their brands. In a preliminary user study, we analyzed users’ gaze behavior to reveal how people interact with the carousels of the multi-list for product comparison. The results show that people explored the position of items in all the carousels, following a pattern that differs from the top-left triangle observed in traditional multi-lists, and they selected items having a fairly good ranking, showing their interest in sustainability and ethical standards.

Validation of the EDUSS Framework for Self-Actualization Based on Transparent User Models: A Qualitative Study

Self-actualization is the process of striving toward full potential and achieving higher goals in one’s life. Originally studied in psychology, this concept has been adopted by various disciplines, including recommender systems, as a means of addressing issues like the filter bubble problem and promoting transparency. In an earlier work, we developed a theoretically-sound framework named EDUSS to systematically design interactive visualizations of transparent user models for self-actualization. We aim in this paper to validate the effectiveness of using the EDUSS framework to support self-actualization. To this end, we implemented interactive visualizations of transparent user interest models designed with the help of the EDUSS framework into the transparent Recommendation and Interest Modeling Application (RIMA). Further, we conducted a qualitative user study (N=10) to investigate the effect of these visualizations in supporting users to achieve self-actualization. Our study showed qualitative evidence validating that applying the EDUSS framework to design systems for self-actualization has the potential to help users reach self-actualization goals to a certain extent.

SESSION: FairUMAP 2023: 6th UMAP Workshop on Fairness in User Modeling, Adaptation, and Personalization

6th Workshop on Fairness in User Modeling, Adaptation, and Personalization (FairUMAP 2023)

Understanding User Perspectives on Sustainability and Fairness in Tourism Recommender Systems

Recommender Systems (RS) are widely used in various domains, including travel and tourism, to provide personalized recommendations for accommodations, activities, and destinations. However, the evaluation of RS has traditionally focused on satisfying the needs of end users, item providers, or the recommendation platform itself without considering the impact on society. Sustainable tourism practices are becoming increasingly important, and a Tourism Recommendation System (TRS) can play a crucial role in promoting sustainability by suggesting sustainable activities and less popular destinations to users.

In this study, we explore the effect of integrating sustainable recommendations into TRS to ensure fairness for society. We conduct a user study utilizing synthetic recommendations to assess user perceptions of sustainable options versus unsustainable options. Our contributions include insights into effective strategies for incorporating sustainable items in recommendations, understanding user reactions to sustainable alternatives, and identifying helpful recommendation elements for users in their decision-making process. Our findings demonstrate that including sustainable options in recommendations can encourage tourists to visit sustainable and less popular areas and help address issues such as overtourism and undertourism in the travel and tourism industry.

SESSION: GMAP 2023: 2nd Workshop on Group Modeling, Adaptation and Personalization

GMAP 2023: 2nd Workshop on Group Modeling, Adaptation and Personalization

Although most existing recommender systems support single users, there are many scenarios where these systems target the needs of groups. Traits such as group mood, emotional contagion, and interpersonal relationships are often ill-defined characteristics, tend to mutate over time, and are usually missing from the systems’ modeling, even though they play an indispensable part in group modeling. Furthermore, producing timely and accurate recommendations for groups that are explainable, fair, and privacy-protecting is a notoriously tricky challenge since group members may have divergent views and needs. The second GMAP workshop aims at bringing together a community of researchers focused on group modeling, adaptation, and personalization. The objective is to explore the challenges and opportunities of developing effective methods and tools to support group decision-making. The workshop, we brought together researchers from several disciplines, including Psychology, Computer Science, and Organizational Behavior, to discuss their latest research and ideas on this topic. It also provided opportunities for participants to share their research and experiences and to collaborate and network with other researchers in this field. The long-term goal is to foster a vibrant and inclusive community of researchers committed to advancing our understanding of group modeling, adaptation, and personalization by bringing together experts from different disciplines and perspectives. Throughout this workshop, we aim to identify critical challenges and opportunities in this area and develop a shared research agenda to guide future work.

Adapting Emotional Support in Teams: Emotional Stability and Productivity

Emotional support is a fundamental social construct for human beings, closely tied to mental and physical wellbeing. In the context of a classroom, teachers’ emotional support has been linked to students’ increased motivation, better learning outcomes, and decreased stress, ultimately representing a protective factor against the development of mental illness. Students often work on projects in teams, and many experience issues with teammates, leading to stress and frustration. However, teachers’ limited time and resources represent a challenge to the provision of effective support to such students. Technology is a possible mediator between teachers and students. By means of online interventions, a conversational agent may collect students’ teamwork experiences and deliver support messages at the same time, providing not only a monitoring tool for teachers but also a source of support to students. This intervention requires conversational agents with a validated framework of effective emotional support messages, adapted to the students’ personalities and experiences. In this paper, the first steps for this intervention are presented. First, a corpus of emotional support statements provided by teachers for students working in teams is collected. Second, these statements are validated in emotional support categories. Third, participants are presented with a situation where they have to provide support to a student rating another one on one aspect of group work: Productivity. We investigate the adaptation of such messages to students’ Emotional Stability and the given rating. Two versions of an algorithm are created based on the results.

An Argumentative Framework for Generating Explainable Group Recommendations

In the context of group recommender systems, explanations strategies have been proposed to improve recommendations perceived fairness, consensus, satisfaction, and to help the group members in the decision-making process. In general, such explanations try to clarify the underlying social chioce-based aggregation strategies used to generate the recommendations. However, results in the literature are conflicting, and the real benefit of such explanations seem to be limited. In this work, we propose a novel approach, which makes use of an argumentative framework built using information about the aspects that are connected to the recommended items. Such framework is used to generate recommendations, and related explanations. We provide a proof of concept on how to generate explanations for the group, as well as specific explanations for the group members, which use the information in the argumentative frameworks to enrich the explanations. Furthermore, we propose privacy-preserving versions for the explanations, as well as a graphical approach based on tag clouds. In future works, we plan to evaluate the quality of the provided recommendations in offline settings, as well as the impact of the proposed explanations in a series of user studies.

CHARM: A Group Recommender ChatBot

Group recommender systems (GRSs) are tools that support a group to find items that the whole group would enjoy experiencing jointly. There are two main lines of research in this field. The first line of research focuses on methods that combine the preferences of individual group members to obtain a group preference model and generate appropriate recommendations. The second line of research is more holistic and aims to support groups in all the phases of their decision-making process. The majority of the approaches of the second type use a simple conversational approach, which is critiquing. However, nowadays people heavily rely on social and chat platforms to make group decisions, and we believe that these platforms could be a valuable mean for building more effective GRSs. To this end, we have designed a framework tool that extends standard chat platforms by augmenting it with a chat-bot. The chat-bot enables the communication between the users on one side and the group recommender agent on the other. Our goal is a new holistic approach to group recommendations that would be the more beneficial than previous proposed conversational approaches. We aim to provide the proposed framework as an open environment for researchers to prototype their own GRSs.

Group Adapted Avatar Recommendations for Exergames

Exergames are a promising way to encourage physical activity in the population. Especially competitive gaming has been shown to boost physical activity during gameplay. However, differences in physical abilities and fitness can lead to anxiety, fear of failure, or frustration. One way to mitigate these inhibitors is to balance the exergaming difficulty between competing players. This paper investigates the expectations and attitudes towards adaptivity in sports games, both in real life and with digital support. To that end, we present a survey with 421 participants investigating the general reaction to group adaptivity in sports games as well as a focus group discussing the reactions to group adaptive avatar recommendations within the game Mario Tennis Aces. Our results show that there is potential for group adaptive exergames to increase engagement, especially for non-sporty and female users, and that the first prototypical implementation was perceived positively regarding fairness and expected physical activity.

Less Can Be More: Exploring Population Rating Dispositions with Partitioned Models in Recommender Systems

In this study, we partition users by rating disposition - looking first at their percentage of negative ratings, and then at the general use of the rating scale. We hypothesize that users with different rating dispositions may use the recommender system differently and therefore the agreement with their past ratings may be less predictive of the future agreement.

We use data from a large movie rating website to explore whether users should be grouped by disposition, focusing on identifying their various rating distributions that may hurt recommender effectiveness. We find that such partitioning not only improves computational efficiency but also improves top-k performance and predictive accuracy. Though such effects are largest for the user-based KNN CF, smaller for item-based KNN CF, and smallest for latent factor algorithms such as SVD.

The Effect of Similarity Metric and Group Size on Outlier Selection & Satisfaction in Group Recommender Systems

Group recommender systems (GRS) are a specific case of recommender systems (RS), where recommendations are constructed to a group of users rather than an individual. GRS has diverse application areas including trip planning, recommending movies to watch together, or music in shared environments. However, due to the lack of large datasets with group decision-making feedback information, or even the group definitions, GRS approaches are often evaluated offline w.r.t. individual user feedback and artificially generated groups. These synthetic groups are usually constructed w.r.t. pre-defined group size and inter-user similarity metric. While numerous variants of synthetic group generation procedures were utilized so far, its impact on the evaluation results was not sufficiently discussed. In this paper, we address this research gap by investigating the impact of various synthetic group generation procedures, namely the usage of different user similarity metrics and the effect of group sizes. We consider them in the context of “outlier vs. majority” groups, where a group of similar users is extended with one or more diverse ones. Experimental results indicate a strong impact of the selected similarity metric on both the typical characteristics of selected outliers as well as the performance of individual GRS algorithms. Moreover, we show that certain algorithms better adapt to larger groups than others.

SESSION: HAAPIE 2023: 8th International Workshop on Human Aspects in Adaptive and Personalized Interactive Environments

HAAPIE 2023: 8th International Workshop on Human Aspects in Adaptive and Personalized Interactive Environments

Nowadays, the profound digital transformation has upgraded the role of the computational system into an intelligent multidimensional communication medium that creates new opportunities, competencies, models and processes. The need for human-centered adaptation and personalization is even more recognizable since it can offer hybrid solutions that could adequately support the rising multi-purpose goals, needs, requirements, activities and interactions of users. The HAAPIE workshop1 embraces the essence of the “human-machine co-existence” and brings together researchers and practitioners from different disciplines to present and discuss a wide spectrum of related challenges, approaches and solutions. In this respect, the seventh edition of HAAPIE includes 2 long papers and 5 short papers.

“Who are you?”: Identifying Young Users from a Single Search Query

As an initial step towards enabling the adaptation of (popular, and widely used) web search environments so that they can better serve children and ease their path towards information discovery, we introduce Recognizing Young Searchers (RYSe). RYSe leverages lexical, syntactical, spelling/punctuation, and vocabulary features that align with the Concrete Operational stage of development (originally identified by Jean Piaget) in an attempt to identify users that are in this stage. The concrete operational stage is commonly associated with children ages 7-11. Findings emerging from our initial empirical exploration using single queries formulated by children and sample queries from adults showcase the feasibility of relying on different cognitive traits inferred from the short text of a single query to distinguish those that are formulated by younger searchers.

Children on ChatGPT Readability in an Educational Context: Myth or Opportunity?

In this work, we present the results of a preliminary exploration aiming to understand whether the use of ChatGPT in an educational context can be an asset to meet the specific needs of the students. In particular, we focus on the possibility of adapting the responses to online inquiries related to the primary school curriculum to meet the expectations of readers with different literacy levels. The analysis of feedback elicited from children (9- to 10-year-olds) in three 4th grade classrooms indicates that ChatGPT can adapt its responses to the 4th grade level. However, it still needs improvement to reach the right level of readability. Outcomes from this work can inspire future research directions involving technologies like ChatGPT to adapt learning paths to suit a broad range of students with varied cognitive skills. The potential of such tools to support teachers in their effort to adapt to individual learning needs is still to be fully exploited.

Driver Model for Take-Over-Request in Autonomous Vehicles

This work presents a driver model for Take-Over-Request (TOR) in autonomous vehicles (AVs) that considers the driver’s mental state and response time during the transition from automatic to manual driving modes. The Fallback Ready User (FRU) is introduced as a key component that defines the minimum attention required from the driver to respond to TORs and system failures. Highly adaptive FRU models are essential for ensuring the safety of AVs. By using non-intrusive methods, such as facial expression tracking, to capture the driver’s mental state and improve AV safety we study shared control between the vehicle and driver during TOR. The paper presents two application scenarios for TOR analysis in ADAVEC system: detecting when the driver is ready for TOR and reacting to unpredictable situations, such as driver sickness or drowsiness. The proposed driver model considers user personalization based on high-level features, long-term changes, and real-time evidence.

Not Facial Expression, nor Fingerprint – Acknowledging Complexity and Context in Emotion Research for Human-Centered Personalization and Adaptation

While research on emotion has emerged as a crucial area in studying this relationship, the use of classical psychological concepts in human emotion detection and sentiment analysis has been challenged by the cognitive sciences and psychology. This paper argues that the uncritical adoption of concepts that overlook the complexity and context of emotions may hinder progress in this field. To overcome this limitation, the theory of constructed emotion is reviewed, which suggests that emotions are not distinct categories but rather dimensions that require dynamic, rather than static, contextualized models. By prioritizing digital wellbeing in emotion studies and acknowledging complexity and context, future research can develop more effective models for emotion detection and sentiment analysis. The aim is to provide valuable insights for researchers seeking to advance our understanding of the relationship between technology and wellbeing for human centered-adaptation and personalization.

Recommender Systems in Continuing Professional Education for Public Transport: Challenges of a Human-Centered Design

Continuous training is an essential building block to avoid workforce shortage in the public transport sector in Germany. However, the personnel requirements in this sector are highly diverse, similar to the education history of the employees. Therefore, more and more specialized continuous training offers arise, which are, on the one hand, more and more personalized but also make it more challenging to find suitable offers for the individual. Specialized recommender systems for this niche application might be a possible solution. This paper presents current work-in-progress results towards such a system and, in particular, the requirements for the recommender systems from the users’ perspective. We conducted guided interviews with industry representatives focusing on the usage-oriented expectations in recommender systems for an online platform for offerings of continuing education in the area of public transport. The resulting usage requirements form the basis for the concluding literature review of recommender systems in the special application domain. The results show that especially the challenges of small communities with limited content and multiple profiles are not sufficiently addressed in the development of recommender systems, such that existing solutions are not applicable in this niche area.

Towards Human-Centric Psychomotor Recommender Systems

Recommender Systems have been developed for years to guide the interaction of the users with systems in very diverse domains where information overload exists aimed to help humans in decision making. In order to better support the humans, the more the system knows about the user, the more useful recommendations the user can receive. In this sense, there is a need to explore which are the intrinsic human aspects that should be taken into account in each case when building the user models that provide the personalization. Moreover, there is a need to define and apply methodologies, guidelines and frameworks to develop this kind of systems in order to tackle the challenges of current artificial intelligence applications including issues such as ethics, transparency, explainability and sustainability. For our research, we have chosen the psychomotor domain. To provide some insights into this problem, in this paper we present the research directions we are exploring to apply a human-centric approach when developing the iBAID (intelligent Basket AID) psychomotor system, which aims to recommend the physical activities and movements to perform when training in basketball, either to improve the technique, to recover from an injury or even to keep active when getting older.

Virtual Reality Health Education to Prevent Musculoskeletal Disorders and Chronic Low Back Pain in Formal and Informal Caregivers.

Formal and informal caregivers are suffering from musculoskeletal disorders and low back pain which develops in practice. Previous research has suggested that health education can help to prevent musculoskeletal disorders and low back pain in caregivers. With the ability to simulate real-world scenarios, healthcare education is experiencing rapid growth in the use of immersive technologies such as Virtual Reality, aiming to enhance lifelong learning for caregivers, with promising results. However, the creation of such technologies has not been well documented, which forces educators to face the same setbacks during this development process. Using a user-centred design approach, which involved 14 medical experts and computer scientists, we designed a mobile Virtual Reality application to enhance learning and reduce work attributions for caregivers. Then we evaluated the system with a total of 30 formal and informal caregivers, documenting that Virtual Reality can be an effective solution for lifelong learning. Through this paper, we explain the process and analysis we run to identify how to create an effective Virtual Reality learning system.

SESSION: PATCH 2023: The 14th International Workshop on Personalized Access to Cultural Heritage

14th International Workshop on Personalized Access to Cultural Heritage (PATCH 2023)

Following the successful series of PATCH workshops, PATCH 2023 will again be the meeting point between state-of-the-art cultural heritage (CH) research and personalization research, focused on those using different types of technology, with emphasis on ubiquitous and adaptive scenarios, to enhance the personal experience in CH sites. The workshop is aimed at bringing together researchers and practitioners who are working on various aspects of cultural heritage and are interested in exploring the potential of state-of-the-art mobile and personalized technology (onsite as well as online) to enhance the CH visiting experience. The expected result of the workshop is a multidisciplinary research agenda that will inform future research directions and, hopefully, forge some research collaborations. This summary provides an overview of the papers that have been accepted for presentation at the workshop and for publication in its proceedings.

A Mobile App Supporting Field Trip Organization for Natural and Cultural Heritage Exploration

Mobile tourist guides have great potential to promote Cultural and Natural Heritage but usually do this from a narrow perspective, such as a single exhibition or museum, failing to provide users with an integrated viewpoint of the resources available in a geographical area. The organization of tourist plans might thus be challenging because of the many information sources to be consulted. Current tourist guides also limit users’ freedom in building custom trips because they almost fully control the itinerary generation process. Moreover, they fail to recognize that cultural and scientific tours might include both the visit to places and the execution of activities aimed at deepening people’s experience through experimental work. This is a limitation, especially for the learning field, which recognizes the importance of practical activities in strengthening students’ knowledge and understanding.

To address this issue, we developed the FieldTripOrganizer application as a model to create mobile tourist guides that support the design of plans suitable for cultural/scientific tourism. FieldTripOrganizer empowers users to design a trip by helping them select Points of Interest and activities that are relevant to the interests and knowledge background of the people who will participate in the tour. Moreover, it simultaneously provides information filtering, automated scheduling, and user-awareness support to let users compose the itinerary from scratch while being informed about the feasibility of the options that can be included without violating its time constraints. We exploited FieldTripOrganizer to present the Cultural and Natural resources provided by the Geodidalab scientific laboratory located in the area of Ivrea (Piedmont, Italy).

CARES: an Inclusive Personalized Touristic System for Autism

People have different interests and cognitive capabilities that should be taken into account when developing technological support for cultural heritage exploration. In this project, we aim to help people with autism to plan a tourist trip by taking into account their interests and their cognitive skills. We plan to personalize the suggestion of touristic places and itineraries taking into account different types of constraints such as temporal and physical ones. Moreover, we aim to adapt the user interface of the system on the basis of the users’ capabilities to deliver the right information, using a proper visualization modality, avoiding information overload. In this way, people will be able to know in advance the plan for the trip and this would reduce their level of stress and anxiety. In this paper, we focus on the first stage of the project, i.e. the qualitative interviews we carried out together with the user requirements for our application.

Citizen Curation Methods for Interpretation and Reflection on Cultural Heritage: Insights from SPICE

Developed in the context of the Horizon 2020 project SPICE, this paper explores methods and tools for citizen curation, with a focus on designing activities that can effectively elicit and motivate citizens to produce meaningful interpretations and reflections on cultural heritage. Additionally, the paper aims to evaluate the potential of such methods and activities in supporting the analysis and reflection processes within the SPICE Interpretation-Reflection loop (IRL). We explore a selection of narrative-based methods for citizen curation to propose an approach for eliciting meaningful stories related to cultural heritage artifacts from citizens. The proposed approach is subsequently applied in a digital co-design workshop, conducted with members of the SPICE consortium, including the five SPICE museum partners. The collected story contributions are thereafter analysed using qualitative methods and tools adapted from narrative inquiry and the study of narrative identity. The findings provide valuable insights into the potential of narrative-based approaches for citizen engagement in the cultural heritage domain and support the development of the SPICE digital tools for promoting reflection, i.e. sentiment analysis, personalization, recommendation systems, and user- and community modeling tools. In SPICE, such tools play a central role in enabling citizens to create representations of themselves, while also cultivating an understanding and appreciation of similarities and differences across citizen groups. In this direction, the paper offers ideas and insights for designing engaging participatory approaches to elicit citizen input. At the same time, it aims to support the development of digital systems and tools in the cultural heritage domain, with the goal of fostering more inclusive and dynamic representations of citizens and citizen groups.

Eyeing the Visitor’s Gaze for Artwork Recommendation

Recommender systems (RSs) are increasingly present in our everyday lives for business and pleasure. The Cultural Heritage domain is no exception. In the research literature, several RSs have been proposed to enhance the fruition of artistic and cultural resources. In this paper, we present some of our research activities aimed at realizing a RS for suggesting personalized itineraries to exhibit and museum visitors. More specifically, we describe the collection and use of eye-tracking data to understand if there are any correlations between the visitors’ gaze patterns and their degree of appreciation of the viewed artworks. If such correlations exist, they could be used as implicit feedback in the recommendation engine. The preliminary results are interesting and encourage us to pursue our research activities.

Personalizing Cultural Heritage Access in a Virtual Reality Exhibition: A User Study on Viewing Behavior and Content Preferences

Leveraging digital technologies, museums now have the opportunity to embrace innovative approaches such as knowledge graphs, virtual reality, and virtual assistants to enhance the preservation and interactive presentation of cultural information. However, despite these advancements, personalizing the museum experience remains a significant challenge. Thus, this paper aims to investigate the necessary elements for offering personalized access to cultural heritage within a VR exhibition. To accomplish this, a user study was conducted to identify user preferences for tailored content descriptions, track user viewing behavior to gauge their interest in a VR exhibition, and determine preferred methods of information gathering. The study involved 31 participants, and the findings are expected to provide valuable insights for designing effective and engaging VR exhibitions that cater to diverse visitor interests.

RECBOT: Virtual Museum navigation through a Chatbot assistant and personalized Recommendations

The trend for digitalization of museums has been on the rise in recent years, as museums seek to make their collections and exhibitions more accessible to a wider audience. This has involved the use of technologies such as virtual and augmented reality, online exhibits, and digital archives. These digital initiatives have allowed museums to reach new audiences and provide immersive experiences that enhance visitors’ engagement with the exhibits. Following this trend, in the current work, we propose a conversational agent that assists remote visitors in accessing a museum’s collection. The proposed architecture includes a chatbot for user interaction that employs Natural Language Processing techniques for understanding the user’s input. To increase visitor engagement, a hybrid recommender system is developed that combines content-based and collaborative-filtering components. The available data is modeled in the form of a Knowledge Graph, which allows for useful insights to be extracted from it.

SculptMate: Personalizing cultural heritage experience using fuzzy weights

Virtual Environment (VE) technology has become increasingly popular in the cultural heritage field, providing new ways to experience and interact with cultural artifacts and sites. By creating immersive and realistic virtual environments, VE technology allows users to explore and engage with cultural heritage in a more dynamic and engaging way. Towards this direction, this study introduces SculptMate, a cutting-edge mobile application that uses advanced personalization features (fuzzy logic) to enhance the appreciation and understanding of sculptures from various eras and artistic styles. The application aims to provide users with an immersive and interactive experience, both within and beyond museum settings, by allowing them to explore and interact with an extensive collection of virtual sculptures from museums and galleries worldwide. The paper's objectives are to investigate the potential of SculptMate, examine the effectiveness of fuzzy logic in personalizing the user experience, and assess the impact of the personalized experience on user engagement and satisfaction. The novelty of this study lies in the utilization of fuzzy logic in VE for personalizing the cultural heritage experience. SculptMate has been evaluated with very promising results.

The Sound of Paintings: Using Citizen Curation to Explore the Cross-Modal Personalization of Museum Experiences

This paper describes the use of Citizen Curation to explore ways in which cross-modal experiences can be used and created by museum visitors. Citizen Curation can be defined as individuals and groups from outside the museum profession engaging in curatorial activities to communicate their own ideas and stories. Previous work has explored how Citizen Curation can be used to broaden the range of voices reflected in the museum, thereby widening its appeal and relevance to new audiences. Recent research suggests that cross-modal experiences, combining visual art with music, can enhance the cultural experience as the visitor simultaneously draws on both what they see and hear. Citizen Curation provides a potential method through which visitors can create and share cross-modal experiences for each other, combining visual art and music. In this paper, we introduce the Deep Viewpoints web application that has previously been used for the Citizen Curation of looking at visual art. We then describe how the application was extended to support two further contexts (i) a musicologist curating experiences that link music to visual art in a museum collection, and (ii) visitors to a museum exhibition experiencing and creating cross-modal experiences. Finally, we reflect on different ways in which technology could be used to support cross-modal museum experiences.

Using Recommendations to Affect Social Change in Cultural Heritage: Should We and How?

This paper suggests a different look at recommender systems - instead of the classical approach - recommending what the user may be interested in (including also diversity and serendipity aspects), taking a societal point of view. We suggest examining the use of recommender systems for social recommendations (defined here) to enact social change. We suggest the use of this technology for encouraging visitors to cultural heritage sites to explore diversity of opinions and perspectives related to artifacts and events and to reflect - implementing an interpretation-reflection loop. Visitors are encouraged not only to explore diversity of content but to contribute their own perspective, so others can explore it in the future. We look at this from three perspectives: goals, methods and evaluation.

Virtual Archaeology in a Multi-platform and Multi-lingual Setting

Virtual archaeology has been an active field of investigation in recent decades when imaging techniques have been paired with physical excavations. Also, in recent times, virtual environments have greatly contributed to data interpretation, with applications to scientific dissemination for wide audiences and to scholar investigations that reproduce the excavation field. Underlying the virtual archaeology applications are comprehensive semantic databases that describe both the excavation processes, with the stratigraphic units and the archaeological sites, and the findings with related interpretations. This paper presents BeA-ViR, an application for virtual archaeology that is devoted to general audiences and multi-disciplinary scholars. The archaeological excavation concerns a site in Japan with related multicultural and multilingual issues. BeA-ViR is deployed in three platforms: desktop, CAVE, and web browser. It has been conceived to be effective for large audiences as well as for specialized scholars. It relies on a comprehensive database that merged archaeological and archaeometric knowledge.