PUBLICACIONES

We introduce the Bias Network Approach (BNA) as a sociotechnical method for AI developers to identify, map, and relate biases across the AI development process. This approach addresses the limitations of what we call the "isolationist approach to AI bias," a trend in AI literature where biases are seen as separate occurrences linked to specific stages in an AI pipeline. Dealing with these multiple biases can trigger a sense of excessive overload in managing each potential bias individually or promote the adoption of an uncritical approach to understanding the influence of biases in developers’ decision-making. The BNA fosters dialogue and a critical stance among developers, guided by external experts, using graphical representations to depict biased connections. To test the BNA, we conducted a pilot case study on the "waiting list” project, involving a small AI developer team creating a healthcare waiting list NPL model in Chile. The analysis showed promising findings: (i) the BNA aids in visualizing interconnected biases and their impacts, facilitating ethical reflection in a more accessible way; (ii) it promotes transparency in decision-making throughout AI development; and (iii) more focus is necessary on professional biases and material limitations as sources of bias in AI development.

A current problem in the philosophy of neuroscience consists in determining how to individuate cognitive capacities using neurobiological evidence. One recent proposal grounded on fundamental insights from mechanistic philosophy is an iterative strategy that cycles between neural mechanisms and cognitive capacities, using the former to individuate the latter and vice versa (Francken et al. Synthese, 200(5):378, 2022). However, this view cannot be applied to a fundamental aspect of research on cognitive capacities. Understanding a capacity requires delineating its behavioral profile by identifying the different effects or phenomena associated with it in different task settings. If mechanisms are necessary for integrating these phenomena into a single capacity in a bottom-up way, the iterative cyclic strategy requires a criterion for assessing the across-context identity of the mechanism, which is missing from this view. However, we argue that introducing this criterion turns the strategy into something else. It requires substituting its bottom-up phase with a bi-dimensional stage in which the capacity and its mechanism are individuated simultaneously through the identification of the generalization that connects them and the assessment of how the generalization behaves across contexts.

Publisher:, Link>

ABSTRACT

Medical images are an essential input for the timely diagnosis of pathologies. Despite its wide use in the area, searching for images that can reveal valuable information to support decision-making is difficult and expensive. However, the possibilities that open when making large repositories of images available for search by content are unsuspected. We designed a content-based image retrieval system for medical imaging, which reduces the gap between access to information and the availability of useful repositories to meet these needs. The system operates on the principle of query-by-example, in which users provide medical images, and the system displays a set of related images. Unlike metadata match-driven searches, our system drives content-based search. This allows the system to conduct searches on repositories of medical images that do not necessarily have complete and curated metadata. We explore our system’s feasibility in computational tomography (CT) slices for SARS-CoV-2 infection (COVID-19), showing that our proposal obtains promising results, advantageously comparing it with other search methods.

The affective scaffolding framework underlies differing perspectives on how griefbots affect the grieving process. While some researchers are optimistic, others draw more cautious conclusions. We endorse the view that griefbots are affective scaffolds. However, we draw a pessimistic conclusion about human-griefbot interaction. There are two main views on successful grief: the “Freudian view” and the continuing bonds view. The “Freudian view” paves the way to the pessimistic conclusion. In this paper, we propose a free-energy approach of the “Freudian view”. Predictive processing is often regarded as providing a process theory for the free-energy principle. In this framework, model evidence must be maximised by updating and optimising models. Griefbots are not recommendable since they are scaffolds that obstruct and lengthen the maximisation of model evidence, making models more rigid and poorly adaptable to changing dynamics. Building on this approach, we examine the ethical dimensions of griefbot’s use and design. We argue that there is something inherently morally wrong in the very conception of griefbots. Their functionality predisposes them to cause harm by disrupting the grief process.

Publisher: Revista Bits de Ciencia, Link>

ABSTRACT

Los bots tienen un nefasto efecto en la diseminación de información engañosa o tendenciosa en redes sociales [1]. Su objetivo es amplificar la alcanzabilidad de campañas, transformando artificialmente mensajes en tendencias. Para ello, las cuentas que dan soporte a campañas se hacen seguir por cuentas manejadas por algoritmos. Muchas de las cuentas que siguen a personajes de alta connotación pública son bots, las cuales entregan soporte a sus mensajes con likes y retweets. Cuando estos mensajes muestran un inusitado nivel de reacciones, se transforman en tendencias, lo cual aumenta aún más su visibilidad. Al transformarse en tendencias, su influencia en la red crece, produciendo un fenómeno de bola de nieve.

 

Recent studies on network intrusion detection using deep learning primarily focus on detecting attacks or classifying attack types, but they often overlook the challenge of attributing each attack to its specific source among many potential adversaries (multi-adversary attribution). This is a critical and underexplored issue in cybersecurity. In this study, we address the problem of attacker attribution in complex, multi-step network attack (MSNA) environments, aiming to identify the responsible attacker (e.g., IP address) for each sequence of security alerts, rather than merely detecting the presence or type of attack. We propose a deep learning approach based on Transformer encoders to classify sequences of network alerts and attribute them to specific attackers among many candidates. Our pipeline includes data preprocessing, exploratory analysis, and robust training/validation using stratified splits and 5-fold cross-validation, all applied to real-world multi-step attack datasets from capture-the-flag (CTF) competitions. We compare the Transformer-based approach with a multilayer perceptron (MLP) baseline to quantify the benefits of advanced architectures. Experiments on this challenging dataset demonstrate that our Transformer model achieves near-perfect accuracy (99.98%) and F1-scores (macro and weighted ≈ 99%) in attack attribution, significantly outperforming the MLP baseline (accuracy 80.62%, macro F1 65.05% and weighted F1 80.48%). The Transformer generalizes robustly across all attacker classes, including those with few samples, as evidenced by per-class metrics and confusion matrices. Our results show that Transformer-based models are highly effective for multi-adversary attack attribution in MSNA, a scenario not or under-addressed in the previous intrusion detection systems (IDS) literature. The adoption of advanced architectures and rigorous validation strategies is essential for reliable attribution in complex and imbalanced environments.

The COVID-19 outbreak implied many changes in the daily life of most of the world's population for a long time, prompting severe restrictions on sociality. The Behavioral Immune System (BIS) suggests that when facing pathogens, a psychological mechanism would be activated that, among other things, would generate an increase in prejudice and discrimination towards marginalized groups, including immigrants. This study aimed to test if people tend to enhance their rejection of minorities and foreign groups under the threat of contagious diseases, using the users' attitudes towards migrants in Twitter data from Chile, for pre-pandemic and pandemic contexts. Our results appear to be mostly against the BIS hypothesis, with some faint exceptions, since threatened users increased their tweet production in the pandemic period, compared to empathetic users, but the latter grew in number and also increased the reach of their tweets between the two periods. We also found differences in the use of language between these types of users. Alternative explanations for these results may be context-dependent.

Publisher: arXiv, Link>

ABSTRACT

Current language models are usually trained using a self-supervised scheme, where the main focus is learning representations at the word or sentence level. However, there has been limited progress in generating useful discourse-level representations. In this work, we propose to use ideas from predictive coding theory to augment BERT-style language models with a mechanism that allows them to learn suitable discourse-level representations. As a result, our proposed approach is able to predict future sentences using explicit top-down connections that operate at the intermediate layers of the network. By experimenting with benchmarks designed to evaluate discourse-related knowledge using pre-trained sentence representations, we demonstrate that our approach improves performance in 6 out of 11 tasks by excelling in discourse relationship detection.

Recent cases of forced explantation of neurotechnologies seem to be grounded on a naturalist conception of the body as an entity that cannot have a non-biological object as a proper part. However, this conception has been challenged by functional approaches, according to which if an artifact robustly contributes to the function of a body, it is part of it and should be legally treated as such. Bublitz (2022) argues that a series of problems would result from revising the law to accommodate a functional view and, for this reason, naturalism is the best option. We claim that it is unacceptable to endorse naturalism for purely pragmatic reasons while recognizing that it is theoretically groundless. We argue that contemporary versions of Autopoietic Theory can be used to provide a theoretically sound naturalistic view. We articulate a criterion for the attribution of degrees of bodiliness to any given object, depending on how closely it is related to autopoiesis, and then specify a threshold that defines the degree required to be a part of the body. Crucially, according to our view, only a very restricted set of devices can become body parts, which significantly mitigates the legal problems of body/device hybridization.

The minimax sample complexity of group distributionally robust optimization (GDRO) has been determined up to a \log(K) factor, where K is the number of groups. In this work, we venture beyond the minimax perspective via a novel notion of sparsity that we dub (\lambda, \beta)-sparsity. In short, this condition means that at any parameter \theta, there is a set of at most \beta groups whose risks at \theta all are at least \lambda larger than the risks of the other groups. To find an \epsilon-optimal \theta, we show via a novel algorithm and analysis that the \epsilon-dependent term in the sample complexity can swap a linear dependence on K for a linear dependence on the potentially much smaller \beta. This improvement leverages recent progress in sleeping bandits, showing a fundamental connection between the two-player zero-sum game optimization framework for GDRO and per-action regret bounds in sleeping bandits. We next show an adaptive algorithm which, up to log factors, gets a sample complexity bound that adapts to the best (\lambda, \beta)-sparsity condition that holds. We also show how to get a dimension-free semi-adaptive sample complexity bound with a computationally efficient method. Finally, we demonstrate the practicality of the (\lambda, \beta)-sparsity condition and the improved sample efficiency of our algorithms on both synthetic and real-life datasets.