Fake news has emerged as a pervasive problem within Online Social Networks, leading to a surge of research interest in this area. Understanding the dissemination mechanisms of fake news is crucial in comprehending the propagation of disinformation/misinformation and its impact on users in Online Social Networks. This knowledge can facilitate the development of interventions to curtail the spread of false information and inform affected users to remain vigilant against fraudulent/malicious content. In this paper, we specifically target the Twitter platform and propose a Multivariate Hawkes Point Processes model that incorporates essential factors such as user networks, response tweet types, and user stances as model parameters. Our objective is to investigate and quantify their influence on the dissemination process of fake news. We derive parameter estimation expressions using an Expectation Maximization algorithm and validate them on a simulated dataset. Furthermore, we conduct a case study using a real dataset of fake news collected from Twitter to explore the impact of user stances and tweet types on dissemination patterns. This analysis provides valuable insights into how users are influenced by or influence the dissemination process of disinformation/misinformation, and demonstrates how our model can aid in intervening in this process.

The authors introduce BERTNN (Bidirectional Encoder Representations from Transformers Neural Network), a novel methodology designed to expand affective lexicons, a critical component in sociological research. BERTNN estimates the affective meanings and their distribution for new concepts, bypassing the need for extensive surveys by leveraging their contextual usage in language. The cornerstone of BERTNN is the use of nuanced word embeddings from Bidirectional Encoder Representations from Transformers. BERTNN uniquely encodes words within the framework of synthesized social event sentences, preserving their meaning across actor-behavior-object positions. The model is fine-tuned on the basis of the implied sentiment changes, providing a more refined estimation of affective meanings. BERTNN outperforms previous approaches, setting a new standard in deriving multidimensional affective meanings for novel concepts. It efficiently replicates sentiment ratings that traditionally require extensive survey hours, demonstrating the power of automated modeling in sociological research. The expanded affective lexicons that can be produced with BERTNN cater to shifting cultural meanings and diverse subgroups, demonstrating the potential of computational linguistics to enrich the measurement tools in sociological research. This article underscores the novelty and significance of BERTNN in the broader context of sociological methodology.

There has been a recent surge in studies applying artificial intelligence, most notably machine learning (ML), to questions about environmental sustainability and climate change. Particularly, there has been growing interest in combining ML and life cycle assessment (LCA) as a means to expand the breadth and depth of LCA studies. However, much of the ML-integrated LCA (ML + LCA) work published to date has not considered the uncertainty of ML modeling. This study explores the application of ML techniques for use in LCA with careful focus on propagating and managing uncertainty. An existing open-access ML + LCA model of hydrothermal biomass treatment was selected as a case study. The benchmark model was rebuilt, and four different uncertainty treatments (cases) were evaluated: (I) no uncertainty analysis, (II) uncertainty analysis for ML only, (III) uncertainty analysis for LCA only (via Monte Carlo), and (IV) uncertainty analysis for both ML and LCA. Three techniques were used to generate ML-based parameter predictions with corresponding prediction intervals (PIs), including Random Forest with Quantile Regression, Artificial Neural Network with Monte Carlo dropout, and Natural Gradient Boosting (NGBoost). Relevant metrics were then used to evaluate the resulting PIs, including Prediction Interval Coverage Probability (PICP), Mean Prediction Interval Width (MPIW), and overall interval score. Results indicate that NGBoost outperforms other ML techniques for the case study of interest, achieving acceptable ML validity measures and much narrower PIs than the other evaluated techniques. For Case I (no uncertainty), the single value Global Warming Potential (GWP) estimate was approximately 1 kg CO2eq/Mt feedstock dry weight (DW). For Case II (ML uncertainty only), the interpercentile interval for GWP ranged from -170 to 152 kg CO2eq/Mt DW. Case III (LCA uncertainty only) exhibited an even wider interpercentile interval (-225, 288). Case IV (ML and LCA uncertainty) exhibited the widest interpercentile interval (-257, 344). These results highlight how accounting for multiple sources of uncertainty within an ML + LCA framework can influence interpretation of overall modeling results. Furthermore, the findings emphasize the importance of aligning the choice of ML technique with a specific model structure to ensure that results are meaningful for highstakes decision-making.

A model based on the cluster process representation of the self-exciting process model is derived to allow for variation in the excitation effects for terrorist events in a self-exciting or cluster process model. The model's derivation and implementation details are given and applied to data from the Global Terrorism Database (National Consortium for the Study of Terrorism and Responses to Terrorism (START), 2015) from 2000 to 2013. Results regarding the practical interpretation and implications for a theoretical model paralleling existing criminological theory are discussed.

Law enforcement agencies are tasked with crime prevention and crime reduction under limited resources. Having an accurate temporal estimate of the crime rate would be valuable to achieve such a goal. However, estimation is usually complicated by the interval censored nature of crime data. We cast the problem of intensity estimation as a Poisson regression using an EM algorithm to estimate the parameters. Two special penalties are added that provide smoothness over the time of day and day of week. This approach provides accurate intensity estimates and can also uncover day of week clusters that share the same intensity patterns. Both simulated and real crime data gathered from the city of Cincinnati and the city of Dallas are used to demonstrate the effectiveness of the proposed model.

Purpose: The 2020 ISHLT consensus statement on pediatric donor heart acceptability suggested donor characteristics have minimal impact on post-transplant survival. Nevertheless, in the U.S., over 90% of all offers to pediatric candidates are refused and almost 40% of pediatric donor hearts are ultimately discarded. We used machine learning (ML) to evaluate the importance of different variable types, including donor variables, in predicting pediatric heart transplant survival.

Background: Organ procurement organizations (OPOs) are responsible for the medical management of organ donors. Given the variability in pediatric donor heart utilization among OPOs, we examined factors that may explain this variability, including differences in donor medical management, organ quality, and candidate factors. Methods: The Organ Procurement and Transplant Network database was queried for pediatric (<18 years) heart donors and candidates receiving pediatric donor heart offers from 2010 to 2019. OPOs were stratified by pediatric donor heart utilization rate, and the top and bottom quintiles were compared based on donor management strategies and outcomes. A machine learning algorithm, combining 11 OPO, donor, candidate, and offer variables, was used to determine factors most predictive of whether a heart offer is accepted. Results: There was no clinically significant difference between the top and bottom quintile OPOs in baseline donor characteristics, distance between donor and listing center, management strategies, or organ quality. Machine learning modeling suggested neither OPO donor management nor cardiac function is the primary driver of whether an organ is accepted. Instead, number of prior donor offer refusals and individual listing center receiving the offer were two of the most predictive variables of organ acceptance. Conclusions: OPO clinical practice variation does not seem to account for the discrepancy in pediatric donor heart utilization rates among OPOs. Listing center acceptance practice and prior number of donor refusals seem to be the important drivers of heart utilization and may at least partially account for the variation in OPO heart utilization rates given the regional association between OPOs and listing centers.

Accurate time series forecasting is critical in various fields, including resource allocation and crime prevention. While traditional approaches often focus on continuous data, count data forecasting, especially with partially observed (censored) data, remains challenging. This paper introduces DeepCensored, a novel deep learning-based framework that combines the Expectation-Maximization (EM) algorithm with deep neural networks to deliver robust probabilistic forecasts. DeepCensored naturally handles interval-censored event data, where exact event times are unknown but fall within a specific interval. Through extensive simulations and real-world crime data analysis, our method significantly outperforms traditional forecasting approaches, reducing the Mean Absolute Error (MAE) by 50% and better detection of emerging crime trends. These results highlight DeepCensored's potential to provide actionable insights for law enforcement and public safety by predicting crime intensity and enabling resource-efficient policing strategies.

Breakthrough technologies have the potential to disrupt markets and society. Anticipating such disruptions is crucial for policymakers, investors, and businesses in being proactive with regard to regulatory policies and in allocating resources effectively. This project aims to develop an analytical approach to identify companies that will lead in developing breakthrough technologies. The analysis focuses on the semiconductor industry, which has seen rapid growth in recent decades, surging from $139 billion in revenue in 2001 to $573.5 billion in 2022. Our systematic approach to predicting technological disruption in the semiconductor industry involves leveraging a combination of quantitative company data, human-centric elements, and feature engineering. Data was collected on 244 private semiconductor companies between 2012 and 2018, encompassing information about leadership profiles, research endeavors, media exposure, and financial performance. Two models were developed: a penalized regression model, and a boosted tree model, both aimed at forecasting the probability of a company achieving a valuation exceeding $500 million within five years of its first funding deal. Key variables such as the number of employees, year founded, total invested equity, number of active patents, and country of origin emerged as significant predictors of company success. This paper discusses the performance of our models and explores applying our findings to identify disruptive companies across industries.

BACKGROUND:Recent studies demonstrate high offer decline and organ non-utilization rates are associated with increased pediatric heart transplant waitlist mortality. We sought to determine which donor, candidate, and offer specific variables most importantly influenced these decisions using only data available at the time of each offer. METHODS:Retrospective review of pediatric (<18 years) heart donor offers made to pediatric candidates in the United States between 2010 and 2020. In addition to standard donor, candidate, and offer data available in UNOS, we extracted objective and qualitative valvar and myocardial function data from all available donor echocardiogram reports. RESULTS:During the study period, 5625 pediatric donor hearts produced 30 156 offers to 4905 unique candidates, of which 88.7% of all offers were declined and 39.2% of organs were not utilized by pediatric waitlisted candidates. Of the 60.8% utilized hearts, 89.7% had a 'cumulatively' normal echocardiogram at the time of offer acceptance; 62.9% of hearts not utilized for a pediatric candidate also had a cumulatively normal final echocardiogram. Random forest and logistic regression modeling demonstrated good predictive performance (AUROC ≥0.83) of likelihood to accept when utilizing donor, candidate, and offer specific variables. SHAP variable importance scores demonstrated number of prior offer declines and candidate institution's prior year acceptance rates as the two most important variables influencing offer decisions. CONCLUSIONS:Behavioral economics appear to play a significant role in pediatric heart transplant candidate institutions' acceptance practices, even when considering the arguably healthier pediatric donor population. Removal of prior institution's decisions from DonorNet may help increase donor utilization.

AI technologies have made significant advancements across various sectors, especially healthcare. Although AI algorithms in healthcare showcase remarkable predictive capabilities, apprehensions have emerged owing to errors, biases, and a lack of transparency. These concerns have led to a decline in trust among clinicians and patients, while also posing the risk of further accentuating pre-existing biases against marginalized groups and exacerbating inequities. This paper presents a scenario-based preferences risk register 1 1 Denotes a methodically arranged document or database detailing potential risks linked to particular scenarios or situations. framework for identifying and accounting AI algorithm biases in diagnosing diseases. The framework is demonstrated with a realistic case study on cardiac sarcoidosis. The framework identifies success criteria, initiatives, emergent conditions and the most and least disruptive scenarios. The success criteria align with the National Institute of Standards and Technology AI Risk Management Framework (NIST AI RMF) trustworthy AI characteristics, and the scenarios are based on various statistical/computational bias that causes algorithmic bias. The framework provides valuable guidance for leveraging AI in healthcare, enhancing objective designs, and mitigating risks by adopting a figure of merit to score the initiatives and measuring the disruptive order. By prioritizing transparency, trustworthy AI, and identifying the most and least disruptive scenarios/biases, the framework promotes responsible and effective use of AI technologies in healthcare.

This paper focuses on the EmoWoz dataset, an extension of MultiWOZ that provides emotion labels for the dialogues. MultiWOZ was partitioned initially for another purpose, resulting in a distributional shift when considering the new purpose of emotion recognition. The emotion tags in EmoWoz are highly imbalanced and unevenly distributed across the partitions, which causes sub-optimal performance and poor comparison of models. We propose a stratified sampling scheme based on emotion tags to address this issue, improve the dataset's distribution, and reduce dataset shift. We also introduce a special technique to handle conversation (sequential) data with many emotional tags. Using our proposed sampling method, models built upon EmoWoz can perform better, making it a more reliable resource for training conversational agents with emotional intelligence. We recommend that future researchers use this new partitioning to ensure consistent and accurate performance evaluations.

Introduction:Technical burdens and time-intensive review processes limit the practical utility of video capsule endoscopy (VCE). Artificial intelligence (AI) is poised to address these limitations, but the intersection of AI and VCE reveals challenges that must first be overcome. We identified five challenges to address. Challenge #1: VCE data are stochastic and contains significant artifact. Challenge #2: VCE interpretation is cost-intensive. Challenge #3: VCE data are inherently imbalanced. Challenge #4: Existing VCE AIMLT are computationally cumbersome. Challenge #5: Clinicians are hesitant to accept AIMLT that cannot explain their process.Methods:An anatomic landmark detection model was used to test the application of convolutional neural networks (CNNs) to the task of classifying VCE data. We also created a tool that assists in expert annotation of VCE data. We then created more elaborate models using different approaches including a multi-frame approach, a CNN based on graph representation, and a few-shot approach based on meta-learning.Results:When used on full-length VCE footage, CNNs accurately identified anatomic landmarks (99.1%), with gradient weighted-class activation mapping showing the parts of each frame that the CNN used to make its decision. The graph CNN with weakly supervised learning (accuracy 89.9%, sensitivity of 91.1%), the few-shot model (accuracy 90.8%, precision 91.4%, sensitivity 90.9%), and the multi-frame model (accuracy 97.5%, precision 91.5%, sensitivity 94.8%) performed well.Discussion:Each of these five challenges is addressed, in part, by one of our AI-based models. Our goal of producing high performance using lightweight models that aim to improve clinician confidence was achieved.

Predicting the outcome of a process requires modeling the system dynamic and observing the states. In the context of social behaviors, sentiments characterize the states of the system. Affect Control Theory (ACT) uses sentiments to manifest potential interaction. ACT is a generative theory of culture and behavior based on a three-dimensional sentiment lexicon. Traditionally, the sentiments are quantified using survey data which is fed into a regression model to explain social behavior. The lexicons used in the survey are limited due to prohibitive cost. This paper uses a fine-tuned Bidirectional Encoder Representations from Transformers (BERT) model to develop a replacement for these surveys. This model achieves state-ofthe-art accuracy in estimating affective meanings, expanding the affective lexicon, and allowing more behaviors to be explained.

Disinformation and misinformation, spread over online social media, can be harmful to society. While there are a number of models that predict or explain the dynamic behavior of fake news dissemination over social media, the limited availability of datasets hinders the ability to fully test these models under a wide range of possible conditions. Simulation, however, offers a low-cost approach to evaluate models over many different scenarios. This paper proposes a novel approach for simulating fake news dissemination process on Twitter. Our approach combines a Multivariate Hawkes Point Processes with concepts from Agent-Based models, to generate realistic data that incorporates the core elements of Twitter including user networks, tweet type, user stance toward the news event, and time. The flexible and efficient simulation approach can capture a wide range of realistic behavior through a rich set of tuning parameters. We show how closely the simulated data can replicate real fake news events.

Forecasting plays a critical role in the development of organisational business strategies. Despite a considerable body of research in the area of forecasting, the focus has largely been on the financial and economic outcomes of the forecasting process as opposed to societal benefits. Our motivation in this study is to promote the latter, with a view to using the forecasting process to advance social and environmental objectives such as equality, social justice and sustainability. We refer to such forecasting practices as Forecasting for Social Good (FSG) where the benefits to society and the environment take precedence over economic and financial outcomes. We conceptualise FSG and discuss its scope and boundaries in the context of the Doughnut theory. We present some key attributes that qualify a forecasting process as FSG: it is concerned with a real problem, it is focused on advancing social and environmental goals and prioritises these over conventional measures of economic success, and it has a broad societal impact. We also position FSG in the wider literature on forecasting and social good practices. We propose an FSG maturity framework as the means to engage academics and practitioners with research in this area. Finally, we highlight that FSG: (i) cannot be distilled to a prescriptive set of guidelines, (ii) is scalable, and (iii) has the potential to make significant contributions to advancing social objectives.

Cultural sentiments of a society characterize social behaviors, but modeling sentiments to manifest every potential interaction remains an immense challenge. Affect Control Theory (ACT) offers a solution to this problem. ACT is a generative theory of culture and behavior based on a three-dimensional sentiment lexicon. Traditionally, the sentiments are quantified using survey data which is fed into a regression model to explain social behavior. The lexicons used in the survey are limited due to prohibitive cost. This paper uses a fine-tuned Bidirectional Encoder Representations from Transformers (BERT) model for developing a replacement for these surveys. This model achieves state-of-the-art accuracy in estimating affective meanings, expanding the affective lexicon, and allowing more behaviors to be explained.

Understanding how a society views certain policies, politicians, and events can help shape public policy, legislation, and even a political candidate's campaign. This paper focuses on using aggregated, or interval censored, polling data to estimate the times when the public opinion shifts on the US president's job approval. The approval rate is modelled as a Poisson segmented (joinpoint) regression with the EM algorithm used to estimate the model parameters. Inference on the change points is carried out using BIC based model averaging. This approach can capture the uncertainty in both the number and location of change points. The model is applied to president Trump's job approval rating during 2020. Three primary change points are discovered and related to significant events and statements.

Single-species bacterial colony biofilms often present recurring morphologies that are thought to be of benefit to the population of cells within and are known to be dependent on the self-produced extracellular matrix. However, much remains unknown in terms of the developmental process at the single cell level. Here, we design and implement systematic time-lapse imaging and quantitative analyses of the growth of Bacillus subtilis colony biofilms. We follow the development from the initial deposition of founding cells through to the formation of large-scale complex structures. Using the model biofilm strain NCIB 3610, we examine the movement dynamics of the growing biomass and compare them with those displayed by a suite of otherwise isogenic matrix-mutant strains. Correspondingly, we assess the impact of an incomplete matrix on biofilm morphologies and sessile growth rate. Our results indicate that radial expansion of colony biofilms results from the division of bacteria at the biofilm periphery rather than being driven by swelling due to fluid intake. Moreover, we show that lack of exopolysaccharide production has a negative impact on cell division rate, and the extracellular matrix components act synergistically to give the biomass the structural strength to produce aerial protrusions and agar substrate-deforming ability.