arXiv subjects

Carlos Castillo

Publications and source records attributed to Carlos Castillo.

At least 37 records · Page 2Linked to original sources

arXiv2022

Human Response to an AI-Based Decision Support System: A User Study on the Effects of Accuracy and Bias

Artificial Intelligence (AI) is increasingly used to build Decision Support Systems (DSS) across many domains. This paper describes a series of experiments designed to observe human response to different characteristics of a DSS such as accuracy and bias, particularly the extent to which participants rely on the DSS, and the performance they achieve. In our experiments, participants play a simple online game inspired by so-called "wildcat" (i.e., exploratory) drilling for oil. The landscape has two layers: a visible layer describing the costs (terrain), and a hidden layer describing the reward (oil yield). Participants in the control group play the game without receiving any assistance, while in treatment groups they are assisted by a DSS suggesting places to drill. For certain treatments, the DSS does not consider costs, but only rewards, which introduces a bias that is observable by users. Between subjects, we vary the accuracy and bias of the DSS, and observe the participants' total score, time to completion, the extent to which they follow or ignore suggestions. We also measure the acceptability of the DSS in an exit survey. Our results show that participants tend to score better with the DSS, that the score increase is due to users following the DSS advice, and related to the difficulty of the game and the accuracy of the DSS. We observe that this setting elicits mostly rational behavior from participants, who place a moderate amount of trust in the DSS and show neither algorithmic aversion (under-reliance) nor automation bias (over-reliance).However, their stated willingness to accept the DSS in the exit survey seems less sensitive to the accuracy of the DSS than their behavior, suggesting that users are only partially aware of the (lack of) accuracy of the DSS.

cs.AI↗

arXiv2022

Rewiring What-to-Watch-Next Recommendations to Reduce Radicalization Pathways

Recommender systems typically suggest to users content similar to what they consumed in the past. If a user happens to be exposed to strongly polarized content, she might subsequently receive recommendations which may steer her towards more and more radicalized content, eventually being trapped in what we call a "radicalization pathway". In this paper, we study the problem of mitigating radicalization pathways using a graph-based approach. Specifically, we model the set of recommendations of a "what-to-watch-next" recommender as a d-regular directed graph where nodes correspond to content items, links to recommendations, and paths to possible user sessions. We measure the "segregation" score of a node representing radicalized content as the expected length of a random walk from that node to any node representing non-radicalized content. High segregation scores are associated to larger chances to get users trapped in radicalization pathways. Hence, we define the problem of reducing the prevalence of radicalization pathways by selecting a small number of edges to "rewire", so to minimize the maximum of segregation scores among all radicalized nodes, while maintaining the relevance of the recommendations. We prove that the problem of finding the optimal set of recommendations to rewire is NP-hard and NP-hard to approximate within any factor. Therefore, we turn our attention to heuristics, and propose an efficient yet effective greedy algorithm based on the absorbing random walk theory. Our experiments on real-world datasets in the context of video and news recommendations confirm the effectiveness of our proposal.

cs.CY↗

arXiv2022

A Comparative User Study of Human Predictions in Algorithm-Supported Recidivism Risk Assessment

In this paper, we study the effects of using an algorithm-based risk assessment instrument to support the prediction of risk of criminalrecidivism. The instrument we use in our experiments is a machine learning version ofRiskEval(name changed for double-blindreview), which is the main risk assessment instrument used by the Justice Department ofCountry(omitted for double-blind review).The task is to predict whether a person who has been released from prison will commit a new crime, leading to re-incarceration,within the next two years. We measure, among other variables, the accuracy of human predictions with and without algorithmicsupport. This user study is done with (1)generalparticipants from diverse backgrounds recruited through a crowdsourcing platform,(2)targetedparticipants who are students and practitioners of data science, criminology, or social work and professionals who workwithRiskEval. Among other findings, we observe that algorithmic support systematically leads to more accurate predictions fromall participants, but that statistically significant gains are only seen in the performance of targeted participants with respect to thatof crowdsourced participants. We also run focus groups with participants of the targeted study to interpret the quantitative results,including people who useRiskEvalin a professional capacity. Among other comments, professional participants indicate that theywould not foresee using a fully-automated system in criminal risk assessment, but do consider it valuable for training, standardization,and to fine-tune or double-check their predictions on particularly difficult cases.

cs.CY↗

arXiv2022

Diversity in the Music Listening Experience: Insights from Focus Group Interviews

Music listening in today's digital spaces is highly characterized by the availability of huge music catalogues, accessible by people all over the world. In this scenario, recommender systems are designed to guide listeners in finding tracks and artists that best fit their requests, having therefore the power to influence the diversity of the music they listen to. Albeit several works have proposed new techniques for developing diversity-aware recommendations, little is known about how people perceive diversity while interacting with music recommendations. In this study, we interview several listeners about the role that diversity plays in their listening experience, trying to get a better understanding of how they interact with music recommendations. We recruit the listeners among the participants of a previous quantitative study, where they were confronted with the notion of diversity when asked to identify, from a series of electronic music lists, the most diverse ones according to their beliefs. As a follow-up, in this qualitative study we carry out semi-structured interviews to understand how listeners may assess the diversity of a music list and to investigate their experiences with music recommendation diversity. We report here our main findings on 1) what can influence the diversity assessment of tracks and artists' music lists, and 2) which factors can characterize listeners' interaction with music recommendation diversity.

cs.HC↗

arXiv2022

Modeling and mitigating human annotation errors to design efficient stream processing systems with human-in-the-loop machine learning

High-quality human annotations are necessary for creating effective machine learning-driven stream processing systems. We study hybrid stream processing systems based on a Human-In-The-Loop Machine Learning (HITL-ML) paradigm, in which one or many human annotators and an automatic classifier (trained at least partially by the human annotators) label an incoming stream of instances. This is typical of many near-real-time social media analytics and web applications, including annotating social media posts during emergencies by digital volunteer groups. From a practical perspective, low-quality human annotations result in wrong labels for retraining automated classifiers and indirectly contribute to the creation of inaccurate classifiers. Considering human annotation as a psychological process allows us to address these limitations. We show that human annotation quality is dependent on the ordering of instances shown to annotators and can be improved by local changes in the instance sequence/order provided to the annotators, yielding a more accurate annotation of the stream. We adapt a theoretically-motivated human error framework of mistakes and slips for the human annotation task to study the effect of ordering instances (i.e., an "annotation schedule"). Further, we propose an error-avoidance approach to the active learning paradigm for stream processing applications robust to these likely human errors (in the form of slips) when deciding a human annotation schedule. We support the human error framework using crowdsourcing experiments and evaluate the proposed algorithm against standard baselines for active learning via extensive experimentation on classification tasks of filtering relevant social media posts during natural disasters.

cs.HC↗

arXiv2021

Exposure Inequality in People Recommender Systems: The Long-Term Effects

People recommender systems may affect the exposure that users receive in social networking platforms, influencing attention dynamics and potentially strengthening pre-existing inequalities that disproportionately affect certain groups. In this paper we introduce a model to simulate the feedback loop created by multiple rounds of interactions between users and a link recommender in a social network. This allows us to study the long-term consequences of those particular recommendation algorithms. Our model is equipped with several parameters to control (i) the level of homophily in the network, (ii) the relative size of the groups, (iii) the choice among several state-of-the-art link recommenders, and (iv) the choice among three different user behavior models, that decide which recommendations are accepted or rejected. Our extensive experimentation with the proposed model shows that a minority group, if homophilic enough, can get a disproportionate advantage in exposure from all link recommenders. Instead, when it is heterophilic, it gets under-exposed. Moreover, while the homophily level of the minority affects the speed of the growth of the disparate exposure, the relative size of the minority affects the magnitude of the effect. Finally, link recommenders strengthen exposure inequalities at the individual level, exacerbating the "rich-get-richer" effect: this happens for both the minority and the majority class and independently of their level of homophily.

cs.SI↗

arXiv2021

Perceptions of Diversity in Electronic Music: the Impact of Listener, Artist, and Track Characteristics

Shared practices to assess the diversity of retrieval system results are still debated in the Information Retrieval community, partly because of the challenges of determining what diversity means in specific scenarios, and of understanding how diversity is perceived by end-users. The field of Music Information Retrieval is not exempt from this issue. Even if fields such as Musicology or Sociology of Music have a long tradition in questioning the representation and the impact of diversity in cultural environments, such knowledge has not been yet embedded into the design and development of music technologies. In this paper, focusing on electronic music, we investigate the characteristics of listeners, artists, and tracks that are influential in the perception of diversity. Specifically, we center our attention on 1) understanding the relationship between perceived diversity and computational methods to measure diversity, and 2) analyzing how listeners' domain knowledge and familiarity influence such perceived diversity. To accomplish this, we design a user-study in which listeners are asked to compare pairs of lists of tracks and artists, and to select the most diverse list from each pair. We compare participants' ratings with results obtained through computational models built using audio tracks' features and artist attributes. We find that such models are generally aligned with participants' choices when most of them agree that one list is more diverse than the other, while they present a mixed behaviour in cases where participants have little agreement. Moreover, we observe how differences in domain knowledge, familiarity, and demographics can influence the level of agreement among listeners, and between listeners and diversity metrics computed automatically.

cs.IR↗

arXiv2021

SciClops: Detecting and Contextualizing Scientific Claims for Assisting Manual Fact-Checking

This paper describes SciClops, a method to help combat online scientific misinformation. Although automated fact-checking methods have gained significant attention recently, they require pre-existing ground-truth evidence, which, in the scientific context, is sparse and scattered across a constantly-evolving scientific literature. Existing methods do not exploit this literature, which can effectively contextualize and combat science-related fallacies. Furthermore, these methods rarely require human intervention, which is essential for the convoluted and critical domain of scientific misinformation. SciClops involves three main steps to process scientific claims found in online news articles and social media postings: extraction, clustering, and contextualization. First, the extraction of scientific claims takes place using a domain-specific, fine-tuned transformer model. Second, similar claims extracted from heterogeneous sources are clustered together with related scientific literature using a method that exploits their content and the connections among them. Third, check-worthy claims, broadcasted by popular yet unreliable sources, are highlighted together with an enhanced fact-checking context that includes related verified claims, news articles, and scientific papers. Extensive experiments show that SciClops tackles sufficiently these three steps, and effectively assists non-expert fact-checkers in the verification of complex scientific claims, outperforming commercial fact-checking systems.

cs.CL↗

arXiv2021

Predicting Early Dropout: Calibration and Algorithmic Fairness Considerations

In this work, the problem of predicting dropout risk in undergraduate studies is addressed from a perspective of algorithmic fairness. We develop a machine learning method to predict the risks of university dropout and underperformance. The objective is to understand if such a system can identify students at risk while avoiding potential discriminatory biases. When modeling both risks, we obtain prediction models with an Area Under the ROC Curve (AUC) of 0.77-0.78 based on the data available at the enrollment time, before the first year of studies starts. This data includes the students' demographics, the high school they attended, and their admission (average) grade. Our models are calibrated: they produce estimated probabilities for each risk, not mere scores. We analyze if this method leads to discriminatory outcomes for some sensitive groups in terms of prediction accuracy (AUC) and error rates (Generalized False Positive Rate, GFPR, or Generalized False Negative Rate, GFNR). The models exhibit some equity in terms of AUC and GFNR along groups. The similar GFNR means a similar probability of failing to detect risk for students who drop out. The disparities in GFPR are addressed through a mitigation process that does not affect the calibration of the model.

cs.LG↗

arXiv2021

Affirmative Action Policies for Top-k Candidates Selection, With an Application to the Design of Policies for University Admissions

We consider the problem of designing affirmative action policies for selecting the top-k candidates from a pool of applicants. We assume that for each candidate we have socio-demographic attributes and a series of variables that serve as indicators of future performance (e.g., results on standardized tests). We further assume that we have access to historical data including the actual performance of previously selected candidates. Critically, performance information is only available for candidates who were selected under some previous selection policy. In this work we assume that due to legal requirements or voluntary commitments, an organization wants to increase the presence of people from disadvantaged socio-demographic groups among the selected candidates. Hence, we seek to design an affirmative action or positive action policy. This policy has two concurrent objectives: (i) to select candidates who, given what can be learnt from historical data, are more likely to perform well, and (ii) to select candidates in a way that increases the representation of disadvantaged socio-demographic groups. Our motivating application is the design of university admission policies to bachelor's degrees. We use a causal model as a framework to describe several families of policies (changing component weights, giving bonuses, and enacting quotas), and compare them both theoretically and through extensive experimentation on a large real-world dataset containing thousands of university applicants. Our paper is the first to place the problem of affirmative-action policy design within the framework of algorithmic fairness. Our empirical results indicate that simple policies could favor the admission of disadvantaged groups without significantly compromising on the quality of accepted candidates.

cs.CY↗

arXiv2021

Intersectional Affirmative Action Policies for Top-k Candidates Selection

We study the problem of selecting the top-k candidates from a pool of applicants, where each candidate is associated with a score indicating his/her aptitude. Depending on the specific scenario, such as job search or college admissions, these scores may be the results of standardized tests or other predictors of future performance and utility. We consider a situation in which some groups of candidates experience historical and present disadvantage that makes their chances of being accepted much lower than other groups. In these circumstances, we wish to apply an affirmative action policy to reduce acceptance rate disparities, while avoiding any large decrease in the aptitude of the candidates that are eventually selected. Our algorithmic design is motivated by the frequently observed phenomenon that discrimination disproportionately affects individuals who simultaneously belong to multiple disadvantaged groups, defined along intersecting dimensions such as gender, race, sexual orientation, socio-economic status, and disability. In short, our algorithm's objective is to simultaneously: select candidates with high utility, and level up the representation of disadvantaged intersectional classes. This naturally involves trade-offs and is computationally challenging due to the the combinatorial explosion of potential subgroups as more attributes are considered. We propose two algorithms to solve this problem, analyze them, and evaluate them experimentally using a dataset of university application scores and admissions to bachelor degrees in an OECD country. Our conclusion is that it is possible to significantly reduce disparities in admission rates affecting intersectional classes with a small loss in terms of selected candidate aptitude. To the best of our knowledge, we are the first to study fairness constraints with regards to intersectional classes in the context of top-k selection.

cs.CY↗

arXiv2020

A Note on the Significance Adjustment for FA*IR with Two Protected Groups

In this report we provide an improvement of the significance adjustment from the FA*IR algorithm of Zehlike et al., which did not work for very short rankings in combination with a low minimum proportion $p$ for the protected group. We show how the minimum number of protected candidates per ranking position can be calculated exactly and provide a mapping from the continuous space of significance levels ($α$) to a discrete space of tables, which allows us to find $α_c$ using a binary search heuristic.

cs.IR↗

arXiv2020

Social Media Alerts can Improve, but not Replace Hydrological Models for Forecasting Floods

Social media can be used for disaster risk reduction as a complement to traditional information sources, and the literature has suggested numerous ways to achieve this. In the case of floods, for instance, data collection from social media can be triggered by a severe weather forecast and/or a flood prediction. By way of contrast, in this paper we explore the possibility of having an entirely independent flood monitoring system which is based completely on social media, and which is completely self-activated. This independence and self-activation would bring increased robustness, as the system would not depend on other mechanisms for forecasting. We observe that social media can indeed help in the early detection of some flood events that would otherwise not be detected until later, albeit at the cost of many false positives. Overall, our experiments suggest that social media signals should only be used to complement existing monitoring systems, and we provide various explanations to support this argument.

cs.SI↗

arXiv2020

Exploring Artist Gender Bias in Music Recommendation

Music Recommender Systems (mRS) are designed to give personalised and meaningful recommendations of items (i.e. songs, playlists or artists) to a user base, thereby reflecting and further complementing individual users' specific music preferences. Whilst accuracy metrics have been widely applied to evaluate recommendations in mRS literature, evaluating a user's item utility from other impact-oriented perspectives, including their potential for discrimination, is still a novel evaluation practice in the music domain. In this work, we center our attention on a specific phenomenon for which we want to estimate if mRS may exacerbate its impact: gender bias. Our work presents an exploratory study, analyzing the extent to which commonly deployed state of the art Collaborative Filtering(CF) algorithms may act to further increase or decrease artist gender bias. To assess group biases introduced by CF, we deploy a recently proposed metric of bias disparity on two listening event datasets: the LFM-1b dataset, and the earlier constructed Celma's dataset. Our work traces the causes of disparity to variations in input gender distributions and user-item preferences, highlighting the effect such configurations can have on user's gender bias after recommendation generation.

cs.IR↗

arXiv2020

Modeling Islamist Extremist Communications on Social Media using Contextual Dimensions: Religion, Ideology, and Hate

Terror attacks have been linked in part to online extremist content. Although tens of thousands of Islamist extremism supporters consume such content, they are a small fraction relative to peaceful Muslims. The efforts to contain the ever-evolving extremism on social media platforms have remained inadequate and mostly ineffective. Divergent extremist and mainstream contexts challenge machine interpretation, with a particular threat to the precision of classification algorithms. Our context-aware computational approach to the analysis of extremist content on Twitter breaks down this persuasion process into building blocks that acknowledge inherent ambiguity and sparsity that likely challenge both manual and automated classification. We model this process using a combination of three contextual dimensions -- religion, ideology, and hate -- each elucidating a degree of radicalization and highlighting independent features to render them computationally accessible. We utilize domain-specific knowledge resources for each of these contextual dimensions such as Qur'an for religion, the books of extremist ideologues and preachers for political ideology and a social media hate speech corpus for hate. Our study makes three contributions to reliable analysis: (i) Development of a computational approach rooted in the contextual dimensions of religion, ideology, and hate that reflects strategies employed by online Islamist extremist groups, (ii) An in-depth analysis of relevant tweet datasets with respect to these dimensions to exclude likely mislabeled users, and (iii) A framework for understanding online radicalization as a process to assist counter-programming. Given the potentially significant social impact, we evaluate the performance of our algorithms to minimize mislabeling, where our approach outperforms a competitive baseline by 10.2% in precision.

cs.SI↗

arXiv2020

Conflict and Cooperation: AI Research and Development in terms of the Economy of Conventions

Artificial Intelligence (AI) and its relation with societies is increasingly becoming an interesting object of study from the perspective of sociology and other disciplines. Theories such as the Economy of Conventions (EC) are usually applied in the context of interpersonal relations but there is still a clear lack of studies around how this and other theories can shed light on interactions between human an autonomous systems. This work is focused into studying a preliminary step that is a key enabler for the subsequent interaction between machines and humans: how the processes of researching, designing and developing AI related systems reflect different moral registers, represented by conventions within the EC. Having a better understanding of those conventions guiding the advances in AI is considered as the first and required advance to understand the conventions afterwards reflected by those autonomous systems in the interactions with societies. For this purpose, we develop an iterative tool based on active learning to label a data set from the field of AI and Machine Learning (ML) research and present preliminary results of a supervised classifier trained on these conventions. To further demonstrate the feasibility of the approach, the results are contrasted with a classifier trained on software conventions.

cs.CY↗

arXiv2020

SciLens News Platform: A System for Real-Time Evaluation of News Articles

We demonstrate the SciLens News Platform, a novel system for evaluating the quality of news articles. The SciLens News Platform automatically collects contextual information about news articles in real-time and provides quality indicators about their validity and trustworthiness. These quality indicators derive from i) social media discussions regarding news articles, showcasing the reach and stance towards these articles, and ii) their content and their referenced sources, showcasing the journalistic foundations of these articles. Furthermore, the platform enables domain-experts to review articles and rate the quality of news sources. This augmented view of news articles, which combines automatically extracted indicators and domain-expert reviews, has provably helped the platform users to have a better consensus about the quality of the underlying articles. The platform is built in a distributed and robust fashion and runs operationally handling daily thousands of news articles. We evaluate the SciLens News Platform on the emerging topic of COVID-19 where we highlight the discrepancies between low and high-quality news outlets based on three axes, namely their newsroom activity, evidence seeking and social engagement. A live demonstration of the platform can be found here: http://scilens.epfl.ch.

cs.IR↗

arXiv2020

Towards Data-Driven Affirmative Action Policies under Uncertainty

In this paper, we study university admissions under a centralized system that uses grades and standardized test scores to match applicants to university programs. We consider affirmative action policies that seek to increase the number of admitted applicants from underrepresented groups. Since such a policy has to be announced before the start of the application period, there is uncertainty about the score distribution of the students applying to each program. This poses a difficult challenge for policy-makers. We explore the possibility of using a predictive model trained on historical data to help optimize the parameters of such policies.

cs.CY↗