Artificial Intelligence (AI) systems are technologies impacting our lives. The systems learn from existing datasets that record past human decisions. Their performance is measured in terms of accuracy, precision, and recall for reproducing already-known results. Understanding the system’s rationale is crucial to check for bias and accept such technology. Explainable AI (XAI) is the area devoted to opening the AI black box, and designing guidelines to build explainable AI systems. Nevertheless, it is important to understand the user’s needs for these explanations. This paper presents an investigation of the usefulness of XAI systems in the field of cancer diagnosis from the domain expert’s (oncologist) perspective. The main findings suggest domain experts (1) understood the outcomes of the XAI systems; (2) considered XAI outcomes as informative, rather than explanatory; (3) would like to go beyond the fixed presented perspective; and (4) missed the causal relation that would reveal the system’s rationale.

The elaboration of future narratives in Design Fiction has always been an individual activity, showing the unique vision of those who elaborated them and proposed the diegetic artifacts. Creating narratives collectively became possible with the Participatory Design Fiction methodology, which recommends, however, that participants are in the same physical location. We present a methodology to enable the creation of future narratives and their smart devices by a group of geographically dispersed volunteers, the Participatory Design Fiction Extend. With it, it was possible to train people, without knowledge about Design Fiction, in the elaboration of narratives, as well as about smart devices in the future and diegetic artifacts. Volunteers produced narratives and proposed intelligent devices that were later analyzed looking for similarities, differences and complementary ideas, in the light of the concepts of the collective unconscious, enabling the construction of a unique narrative and an intelligent device that contemplated most of the foreseen functionalities.

Technology users will eventually face death in the offline world. However, their data may survive for a long time, recorded in a wide variety of digital services and devices. This digital legacy must be handled and legally abide by the owners' wishes, which may vary from inheritance to deletion. By carrying out a literature review on the subject, the authors noticed there is a lack of user studies on these issues. We, therefore, conducted a focus group with technology users of different profiles to understand the volunteers' acceptability of this topic. During the group discussions, an unexpected factor was raised: to what extent might a user's fame influence their digital immortalization? This study aims to discuss digital immortality in terms of its acceptability among users and to contribute to the field of Human-Computer Interaction (HCI) with important insights on death in the digital world. In our findings regarding acceptability, users can feel some comfort in interacting with their deceased loved ones. Regarding fame, while influential individuals may not need digital immortality, there are concerns regarding the utilitarian nature of their legacies.

Fake news is a problem for society as some people trust the message and react accordingly, propagating the lie. The consequences vary from a simple laugh to a death sentence. Research has focused on distinguishing between fake and fact through machine learning techniques. Most literature reviews and mappings focus on techniques that identify the fake over the message content. In a complementary way, we investigate fake news detection approaches based on the senders’ behavior on social media. We present a conceptual framework that deep dives into different aspects of the main categories of persuader, clarifier, and gullible actors. We also consider automatic and manual profiles. The results provide an overview of existing works on profile trustworthiness in fake news spread and shed light on possible research directions.

Artificial intelligence (AI) has become part of our daily lives and is commonly used in a wide range of industries. However, most of users have little visibility and knowledge about how AI systems reach a particular decision and their outcomes in the various fields where AI and machine learning are being applied. Many of the algorithms' results cannot be understood and explained in terms of how and why a specific decision was taken. With the advance of artificial intelligence researches, several techniques and approaches have emerged over the last few years, but only a few studies have been produced with regards to the end-user perspective and comprehension of the algorithms outcomes. This paper aims to evaluate the comprehensibility of the explanations from the perspective of the different types of users, aiming to generate confidence and understanding of the results produced by the AI.

Mobile technologies have been providing noteworthy scenarios regarding elderly care, replacing the traditional phone calls and face-to-face caregivers by a myriad of sensors and communication networks supporting detailed activities monitoring. On the other hand, the recent COVID pandemic has been leading governments to adopt contact tracing apps aiming at controlling the virus spreading. However, that adoption has raised many legitimate issues regarding individual privacy, besides those related to the effectiveness of this kind of app. In this context, this paper presents FenceBot, a novel approach for safety monitoring of the elderly when outdoors. The elderly data are kept private whenever in normal conditions, and alerts are sent to relatives only on unsafe or contagious risk situations that would demand verification and actions. The proposed technology was evaluated with a discrete-event simulation and the outcomes provided a solid indication of its effectiveness regarding protection from contagious risks situations.

Predicting stock market behavior is a challenge that has been studied and presented several solutions in the literature. Due to technological advances, methodologies have emerged and allowed new approaches to this problem in recent years. Text mining and sentiment analysis have been widely applied in this area. On the other hand, classic solutions as time series analysis continue to be used alone or with new methods. There is still no literature review of the joint use of these methods. In this way, this study presents a systematic review with 57 selected papers using time series, text mining, and sentiment analysis applied to predict financial stock market behavior. Through this research, it was observed that the use of data from social media and internet sites is a compound source of information, providing a better prediction. However, the selection and combination of these data in a relevant way are still limitations found in the proposed models.

Social isolation and loneliness are problems faced by the elderly that might be aggravated due to quarantine during the coronavirus pandemic. This motivates a search demand for automatic technological solutions, such as chatbots, to address the problem, and establish a collaborative interaction between elderly and conversational agents. We analyze the use of conversational agents and the main types of personalization used during interacting with the elderly. We mapped 53 papers in 5 sources to answer 5 research questions that guided our study. Our findings show these solutions may have a positive impact on elderly's life.

Pharmacovigilance is an area of the healthcare sector responsible for drug tests and monitoring. The entire life cycle of the drug release process follows strong rules and procedures that may require many years (10-15 years) and substantial financial resources (up to one billion dollars). As pharmaceutical companies aim to minimize the time and expenses required to release the new drug, they may underreport adverse drug reactions (ADRs). Some companies may fail to research or report all potential reactions to a medicinal product. ADRs can cause harmful or unpleasant reactions users and put lives at risk. This work aims to discover ADRs from digital social media. The main contribution of this work is the creation of a labeled database and an ontology to improve the pre-processing of tweets for classification. On Twitter there are many posts about studies and dissemination of research and messages of experts on subject that are not spontaneous opinions of drug users, but primary focus of this study is to investigate adverse reactions. This shows that social networks can help in the discovery of adverse events, but there is a need for robust pre-processing. Using Twitter as a source for potential drug reaction detection, a database of 30,000 tweets with mentions of four distinct drugs was investigated, which resulted in finding 5 0 adverse reactions. To improve the accuracy of a classification model for ADRs, a supervision of the task is essential. Starting from a database containing 1,000 posts, 50 opinions were identified as ADRs, which is a rate of 5%. This shows that Twitter is a reliable source for the extraction of ADRs.

The advancement of Information and Communication Technologies has brought numerous facilities and benefits to society. In this environment, surrounded by technologies, data, and personal information, have become an essential and coveted tool for many sectors. In this scenario, where a large amount of data has been collected, stored, and shared, privacy concerns arise, especially when dealing with sensitive data such as health data. The information owner generally has no control over his information, which can bring serious consequences such as increases in health insurance prices or put the individual in an uncomfortable situation with disclosing his physical or mental health. While privacy regulations, like the General Data Protection Regulation (GDPR), make it clear that the information owner must have full control and management over their data, disparities have been observed in most systems and platforms. Therefore, they are often not able to give consent or have control and management over their data. For the users to exercise their right to privacy and have sufficient control over their data, they must know everything that happens to them, where their data is, and where they have been. It is necessary that the entire life cycle, from generation to deletion of data, is managed by its owner. To this end, this article presents an Information Chain of Custody Model based on Blockchain technology, which allows from the traceability of information to the offer of tools that will enable the effective management of data, offering total control to its owner. The result showed that the prototype was very useful in the traceability of the information. With that it became clear the technical feasibility of this research.

Data anonymization methods have been investigated as strategies to provide data privacy and to minimize prejudices against minorities. Recent works on machine learning have shown impressive results in generating new examples of images using Generative Adversarial Networks (GAN). Creating synthetic facial images that resembles a group, but no one in particular seem an interesting strategy to anonymize a face. Nevertheless, the image synthesis process is computationally expensive and requires a very large training dataset to produce a human-acceptable outcome. This paper presents an investigation on the trade-off between the required number of instances (size) and the images’ inter-variability of a training set for using Deep Convolutional Generative Adversarial Network (DCGAN) to generate human-acceptable synthetic images. We report an experiment using the Appen crowdsourcing platform to evaluate the human acceptance of the synthetic images generated by our DCGAN trained using different samples of the CelebA public dataset, taken according to facial attributes, such as cheekbones protuberance, face shape and eyebrow thickness. The results indicate that (1) Facial attributes have a significant effect on diminishing the images’ inter-variability, specially face shape; (2) smaller images’ inter-variability leads to smaller required training dataset and (3) filtering the DCGAN’s training datasets using oval face shape led to a required dataset almost 50% lesser than without filtering for generating images humanly acceptable. We believe this result may help individuals to anonymize their visual data maintaining certain characteristics but keeping their privacy at a lower computational cost.

This work extends current collective intelligence evolutionary algorithms by incorporating a collective-based variation operator. As part of this work, the proposals are compared with state-of-the-art reference-point-based MOEAs: NSGA-II and RNSGA-II. Another primary objective of the work is to deal with a real-world multi-objective instance of the facility location problem. The experimental results validate the proposal. The new collective intelligence MOEA outperformed NSGA-II and R-NSGA-II for complex scenarios.

The analysis of data collected through GPS and/or smartphone logs allows the identification of human mobility patterns. This identification has been the subject of several surveys in recent years. However, certain factors can change these patterns, such as special events of great magnitude (e.g., the Olympic Games) or environmental events (e.g., an increase in air pollutants.) In this paper, we present an investigation into the impact of these events on these patterns. We analyze urban mobility data collected in Beijing between April 2009 and October 2012, we use Artificial Intelligence techniques to identify mobility patterns in these data and then check for changes in patterns during the aforementioned events. The analysis of these data has become an important tool to extract insights and to support the planning of the urban displacement in the day to day and in special events.

Online collaborative environments, such as Reddit, have led to the rise of open and self-organizing communities. However, discussions in these places are often scattered, redundant and voluminous, and do not fully support the deliberation flow. Argument polarization, groupthink, subject bubbles, subgroup dominance and lurking are commonly identified behaviors that negatively impact the quality of online discussions. In addition, open environments allow the congregation of homogeneous groups that feed on pre-existing ideas, creating discussion bubbles and contributing to users taking extreme positions on issues. Evidence from Reddit discussions suggest that, while online collaborative environments provide powerful support for the creation and dissemination of new content, their ability to support efficient debate and idea contribution is questionable. Often, debates include hundreds or thousands of people with heterogeneous behaviors concerning post frequency, size, wording and polarity. Joining an ongoing online debate requires understanding what has happened so far. Many studies have focused on summarizing the discussion content as a way to overcome the message volume, dispersion and redundancy. However, to fully understand an online debate, it is also important to perceive the group's dynamics. The identification of dysfunctional behaviors contributes to new users' sensemaking of the debate as a whole and supports their decision whether it is worth joining the group. Additionally, understanding the group dynamics improves participants behavior and the overall discussion. We analyze the use of content summarization and behavior indicators to understand a discussion.

Data Science has assembled researchers from multiple fields, such as computer science and marketing, to develop methods, algorithms and techniques to discover knowledge from large amounts of data and solve complex problems. The main focus is to find patterns hidden within the data. In general, these patterns must appear frequently to be learned using machine algorithms and reflect the standard, even if tacit, knowledge that guides decisions in the world. The uncovered knowledge may also reflect societal prejudice that will be enforced by the machines. Nevertheless, in many domains, specially when the negative impacts of bad decisions have a high cost, a few instances of patterns within the dataset suffice to warrant further investigation. Considering that decisions will be based on knowledge learned from the data, the challenge lies in determining when fewer appearances of a sequence of events are more important than very frequent patterns. In this context, we have devised an approach that uses a domain ontology to boost these infrequent, but relevant, events. In addition to guiding the search for relevant knowledge, the ontology helps users accept results and further investigate the data. It enables users to create data subsets and views, deriving new attributes from existing ones, guiding the data mining process, and providing a background layer from which even not so frequent patterns stand out and become meaningful. In this paper, we present our ontology-based data discovery approach, a system developed according to it, and preliminary results of a real life application in the oil production domain.

We are on the verge of a major shift in the way we perceive digital life, what may cause a significant impact to the real world. Gradually, through increasing knowledge in the areas of artificial intelligence, big data and machine learning, computers have been emulating deceased human beings and, symbolically, with the aid of technology, have been managing to conquer death. This article seeks to understand and problematize the ways in which digital immortality has manifested itself, particularly through digital memorials, chatbots and avatars. Companies like Facebook, Eter9 and Initiative 2045 allow a continuity of life after the death of its users. We analyze, through technical and philosophical questions, imbricated in this discussion, the implications of this digital immortality and how these issues are seen in a differentiated way with the presence of technology.

Automatic dimensionality reduction in text classification requires large training data sets due to the high dimensionality of the native feature space. However, in several real world multi-label problems, such as highly demanding decision-making scenarios, to manually classify and select features in large document sets is usually unfeasible even by specialist teams. This paper presents CrowdFS a first approach on using collective intelligence techniques to select label specific relevant features from a large document set. An experiment in the context of competitive intelligence for a multinational energy company showed CrowdFS producing better results than an automatic state of the art technique.

3D virtual worlds are typically designed to reproduce the physical world, both in the features of the meeting place and also in how people interact with each other. We try to get a better understanding of technical and behavioral issues found in different meeting environments designed for collaborative work. From the problems that we have found, we proposed the SLMeetingRoom, as a model of environment for work meetings which is composed of a set of components to support basic working activities in Second Life. We also performed a pilot study under four different conditions of communication: face-to-face, videoconference, Second Life without the SLMeetingRoom and Second Life with the SLMeetingRoom. The pilot study pointed out that the SLMeetingRoom model is a promising environment according to the criteria: cognitive effort and sense of presence, however, for tasks completeness and participation level, we could not find statistical evidence to support our research hypotheses.

Good decisions require good understanding of the problem, the context and available options. This knowledge, in general, is dispersed through several, frequently unidentified, individuals. Collecting, filtering and putting together the relevant knowledge pieces to create solutions are difficult tasks. Many domains fit this scenario of distributed knowledge and creative solution's needs including governmental decision-making. Democratic governments constantly need to make sense of population demands to properly make decisions that attend overall will. However, low citizens participation in government decisions through democratic processes is an aspect that defies democracy itself. Large-scale citizens participation is difficult for the lack of technology to collect, filter and synthesize information. To deal with this problem, we developed an intelligent mobile network that allows broad citizens' participation and feasible government action. Previous research indicates cellular phones as the device with the broadest acceptance in all Brazilian social classes. Sending messages and voting by cell phones are broken barriers to consider the device as the natural input interface for citizen-government interaction. Additionally, message summarization and clusterization help government understand the demands. In this paper we present mGov, a system for eliciting and answering citizen demands in a participatory government structure. A prototype has been deployed with good results electronic

This paper presents observations from engineering ontology design sessions. We followed the entire design process of a group of engineers building an ontology for root cause analysis to be applied to petroleum pump failure diagnosis. The group was composed of experts with same background, who work in different locations of the same company. Even though they apparently share similar analysis methods, during meetings discrepancies emerged. Different points of view were argued and decisions were made at the end of each meeting. Given the final product acceptance problems, we look back at each meeting and attempt to identify the reasons that led to this unfortunate and common outcome. Understanding the problem enables us to draw a set of guidelines for empowering design meetings with simple technologic methods for improving group final artifact acceptance; therefore diminishing re-work. Our observations point towards (1) a correlation between designers' behavior and future artifact acceptance; (2) a set of actions that interrupts or brings back group attention; and (3) a metric for evaluating group design meetings. We believe our findings may guide software developers to tools for supporting group design.

In multi-agent systems, blackboard communication provides an easy way for agents to communicate. However, in a distributed architecture, the blackboard is commonly implemented in a single node, that is the "blackboard server". This approach may work well for a small number of agents. Nevertheless, when the number of agents increases, the blackboard server becomes the bottleneck for the system scalability. In this work we propose a novel agent communication system based on distributed-shared mechanisms that allows the implementation of a shared address space on a distributed system, without centralizing the blackboard. Our idea is to distribute the shared data over the nodes and use a message passing sub-system, totally transparent to the agents. We implemented this idea in a multi-agent system based on social laws, and used a distributed-shared memory system to handle the messaging. We created a conflict simulator, and showed that the distributed-shared mechanisms easy the implementation of a truly distributed blackboard.

Electronic democracy should provide information and service for the citizens on the Internet, allowing room for debate, participation and electronic voting. The languages being adopted by mass communication means, especially Reality Shows, are efficient and encourage public participation in decision-making. This paper discusses a citizen-government interaction language intended to facilitate citizen participation in the government's decisions. An e-Democracy Model for people participation through web-based technologies is conceived. This model specifies the syntax of an Democracy Interaction Language, a DemIL. Such language incorporates characteristics of Reality Show Formats, and it is the back-end of a web-interface project in the domain researched. The study of case Participative Budget of Brazil represents the language proposed.

Well offshore petroleum exploration is a risky, but very profitable task. Determining the rock formation of each layer of a given reservoir diminishes the risks of expending a great deal of money drilling dry wells. Since collecting offshore rock samples is difficult and expensive, geologists must decide upon drilling based on indirect measures called well log. Traditional statistical methods have been used to assist this task. Neural networks have also been successfully used. As an alternative fuzzy logic based systems have an extra appeal of intuitive comprehension of some uncertainties. This paper presents a hybrid tool that combines neural networks, fuzzy logic and neuro-fuzzy logic to improve human intuition when analyzing the potential of oil fields.

The use of recommender systems (RS) has become common in several e-commerce Web sites. Collaborative filtering is one of the most popular RS techniques. It infers a user's predilections based on similarity with other users. However, to work correctly, the RS depends on the evaluator's participation with truly opinions. The HYRIWYG mechanism proposes an incentive model to motivate users to contribute with a high number of opinions and, more important, guaranteeing the truthfulness of the information. In this paper, we present HYRIWYG mechanism and discuss it using an empirical study in the movie evaluation domain. Initial results highlight the great potential benefits of using this kind of incentive mechanism.

We propose an agenda planning technique with a built-in incentive mechanism, based on the VCG (Vickrey-Clarke-Groves) method from game theory, to help project managers in the engineering construction industry create a more effective agenda. Preliminary results have shown an improvement in both instrumented and perceived meeting quality.

This paper presents HYRIWYG (How You Rate Influences What You Get), a reputation system applicable to Internet Recommendation Systems (RS). The novelty lies in the incentive mechanism that encourages evaluators to volunteer their true opinion. Honesty is encouraged because rewards are indexed by the quality of the RS's suggestions.