Algorithmic Decision Research Articles

Data Governance for Emerging Technologies: A Conceptual Framework for Managing Blockchain, IoT, and AI

As emerging technologies such as Blockchain, the Internet of Things (IoT), and Artificial Intelligence (AI) continue to reshape industries, the need for robust data governance frameworks has become increasingly critical. These technologies introduce unique challenges, including data privacy concerns, security vulnerabilities, and the complexity of managing vast, decentralized data sets. This paper proposes a conceptual framework for data governance tailored to the specific requirements of Blockchain, IoT, and AI technologies. The framework emphasizes a holistic approach, integrating key governance principles such as transparency, accountability, and compliance with regulatory standards. It also highlights the importance of fostering collaboration between stakeholders, including technologists, legal experts, and policymakers, to create a cohesive governance structure that can adapt to the rapid evolution of these technologies. The proposed framework addresses three core areas: data integrity and quality, security and privacy, and ethical considerations. For Blockchain, the focus is on ensuring the immutability and transparency of records while safeguarding against potential misuse of decentralized data. In the context of IoT, the framework prioritizes the management of data from diverse sources, ensuring interoperability and protecting sensitive information from unauthorized access. For AI, the emphasis is on developing ethical guidelines for data usage, preventing bias in algorithmic decision-making, and maintaining transparency in AI-driven processes. The framework also advocates for the integration of advanced data analytics and machine learning techniques to enhance data governance capabilities, enabling real-time monitoring and predictive insights. Additionally, it underscores the need for continuous training and education for all stakeholders to keep pace with the dynamic nature of emerging technologies. By adopting this comprehensive data governance framework, organizations can mitigate risks, ensure compliance, and harness the full potential of Blockchain, IoT, and AI while maintaining public trust.

Benefits and harms associated with the use of AI-related algorithmic decision-making systems by healthcare professionals: a systematic review

Despite notable advancements in artificial intelligence (AI) that enable complex systems to perform certain tasks more accurately than medical experts, the impact on patient-relevant outcomes remains uncertain. To address this gap, this systematic review assesses the benefits and harms associated with AI-related algorithmic decision-making (ADM) systems used by healthcare professionals, compared to standard care. In accordance with the PRISMA guidelines, we included interventional and observational studies published as peer-reviewed full-text articles that met the following criteria: human patients; interventions involving algorithmic decision-making systems, developed with and/or utilizing machine learning (ML); and outcomes describing patient-relevant benefits and harms that directly affect health and quality of life, such as mortality and morbidity. Studies that did not undergo preregistration, lacked a standard-of-care control, or pertained to systems that assist in the execution of actions (e.g., in robotics) were excluded. We searched MEDLINE, EMBASE, IEEE Xplore, and Google Scholar for studies published in the past decade up to 31 March 2024. We assessed risk of bias using Cochrane's RoB 2 and ROBINS-I tools, and reporting transparency with CONSORT-AI and TRIPOD-AI. Two researchers independently managed the processes and resolved conflicts through discussion. This review has been registered with PROSPERO (CRD42023412156) and the study protocol has been published. Out of 2,582 records identified after deduplication, 18 randomized controlled trials (RCTs) and one cohort study met the inclusion criteria, covering specialties such as psychiatry, oncology, and internal medicine. Collectively, the studies included a median of 243 patients (IQR 124-828), with a median of 50.5% female participants (range 12.5-79.0, IQR 43.6-53.6) across intervention and control groups. Four studies were classified as having low risk of bias, seven showed some concerns, and another seven were assessed as having high or serious risk of bias. Reporting transparency varied considerably: six studies showed high compliance, four moderate, and five low compliance with CONSORT-AI or TRIPOD-AI. Twelve studies (63%) reported patient-relevant benefits. Of those with low risk of bias, interventions reduced length of stay in hospital and intensive care unit (10.3 vs. 13.0 days, p=0.042; 6.3 vs. 8.4 days, p=0.030), in-hospital mortality (9.0% vs. 21.3%, p=0.018), and depression symptoms in non-complex cases (45.1% vs. 52.3%, p=0.03). However, harms were frequently underreported, with only eight studies (42%) documenting adverse events. No study reported an increase in adverse events as a result of the interventions. The current evidence on AI-related ADM systems provides limited insights into patient-relevant outcomes. Our findings underscore the essential need for rigorous evaluations of clinical benefits, reinforced compliance with methodological standards, and balanced consideration of both benefits and harms to ensure meaningful integration into healthcare practice. This study did not receive any funding.

Do FinTech algorithms reduce gender inequality in banks loans? A quantitative study from the USA

ABSTRACT The potential of FinTech algorithms to decrease gender bias in credit decisions is limited by the impartiality of the data used to train them. If the data is partial or biased, the algorithmic decision-making process may also be discriminatory, exacerbating existing inequalities. In this study, the effect of FinTech Firms on reducing gender inequality in bank loans in the USA is examined using a loan application from 60 U.S. banks from 2012 to 2022. We use a two-step system GMM approach to estimate the effect of FinTech algorithms on gender bias in credit decisions, focusing on female loan approval rates. Our results show that by controlling the other factor, banks with credit algorithms significantly increased the loan approval rates and thus reduced gender inequality in bank loans. Specifically, the female loan approval rates increased by 8% after banks adopted FinTech algorithms. We also find that the effect is more substantial for banks with higher baseline gender bias in credit decisions. We also performed the Difference in Difference analysis to analyse the policy shocks and FinTech adoption on bank loans’ gender inequality. Results of the study show that FinTech adoption and policy implications have significantly increased the loans for female borrowers. Our findings suggest that FinTech algorithms can potentially mitigate gender bias in credit decisions and promote gender equality in financial services.

CAN SYMBOLIC COMPUTATION AND FORMALIST SYSTEMS ENHANCE MATH EDUCATION WITH ARTIFICIAL INTELLIGENCE?

In recent years, a solution developed using deep learning methods has been used to solve difficult problems in a field. The capability of deep learning models is that they require large and heavily sampled data sets. Computer Algebra Systems developed over time have made significant progress, especially in the field of symbolic mathematics solutions solved by machine learning. It is a persistent problem how appropriate it is to use such formal systems in some aspects of algorithmic decision-making. In this paper, we discussed the suitability of artificial intelligence applications to formal propositions by evaluating a deep learning study conducted especially in the field of symbolic mathematics and Math education. Symbolic computation systems have a strong potential for enhancing math education. Furthermore, within the framework of the Incompleteness Theorem, to show why the construction of a mathematical grammar is not a complete solution for Mathematics education systems.

Public Policies in Data Security in the European Union: A Literature Review

This literature review explores the multifaceted development of public policies in data security within the European Union. It synthesizes findings from fourteen key academic studies and policy analyses, identifying challenges such as regulatory fragmentation, ethical governance, technological uncertainties with blockchain, and the pursuit of digital sovereignty. The analysis reveals that harmonized regulations are needed to bridge gaps across diverse sectors like finance, aviation, and critical infrastructure. Ethical considerations surrounding AI, algorithmic decision-making, and data sharing emerge as essential for fostering public trust. The review also examines institutional dynamics, highlighting the roles of ENISA, national agencies, and EU legislative initiatives like the Digital Operational Resilience Act. It discusses how political negotiations and sector-specific needs complicate the path toward a unified cybersecurity framework. The synthesis underscores the importance of balancing innovation with compliance, safeguarding citizens' rights while promoting technological advancement. Ultimately, the review advocates for a more integrated, ethical, and resilient policy landscape and suggests areas for future research, emphasizing collaboration among EU institutions, member states, industry stakeholders, and researchers to address emerging cyber threats and achieve digital sovereignty.

Advanced Machine Learning – A Comprehensive Survey and New Research Directions

This paper presents a comprehensive survey of advanced machine learning techniques, focusing on recent innovations and emerging research directions. As machine learning continues to evolve, significant strides have been made in areas such as supervised, unsupervised, and reinforcement learning, as well as deep learning methodologies. This survey reviews the latest advancements in these domains, highlighting novel algorithms, hybrid models, and approaches designed to enhance scalability and efficiency. We also explore real-world applications and case studies that illustrate the impact of these advancements across various fields, including healthcare, finance, and robotics. By synthesizing current knowledge, this paper identifies key challenges and gaps in the existing literature, proposing new research directions that hold promise for pushing the boundaries of machine learning. Future research areas include the development of more robust and interpretable models, the integration of machine learning with emerging technologies, and the exploration of ethical considerations in algorithmic decisionmaking. This survey aims to provide a valuable resource for researchers and practitioners seeking to understand the state-of-the-art in machine learning and to stimulate further investigation into these promising new directions.

Algorithms and the "Anti-Preference": A Quantitative Investigation of “Reaching the Wrong Audience” on TikTok

This paper provides an empirical hypothesis test and partial verification of the “algorithmic folk theory” of “reaching the wrong audience.” This user folk theory claims that content posted to TikTok can sometimes become sequestered to a hostile audience, resulting in a sharp influx of harassment and oppositional comments. To test this theory, this paper employs a graphical analysis to identify trends in interaction rates, comments over time, and comment sentiment. The data collected consisted of 1,455 posts and 454,540 comments, which were then evaluated using a natural language processing (NLP) sentiment analysis tool for a total of 297,009,882 effective observations of viewer sentiment response. Using this data, this paper employs a time-series analysis to identify a “resurrecting” post behavior characterized by a sudden increase in engagement of an otherwise “dead” TikTok post as far as ten months after the content’s initial post date. Further, the findings highlighted how this “resurgent behavior” would commonly occur when the sudden influx of engagement contained either distinctly positive or negative comment sentiment. These findings suggest the existence of “audience sentiment sequestering,” explained as the algorithmic restriction of viewership to a specific audience type and a core mechanism of the user folk theory of “reaching the wrong audience.” Lastly, this paper proposes a new theoretical algorithmic phenomenon, the anti-preference theory, to explain why automated algorithmic decision-making may cause a user’s content to “reach the wrong audience” and remain stuck there. This theory suggests that the recommendation algorithm implemented on TikTok is impartial to the positive or negative sentiment of a viewer’s comment but still susceptible to the user’s propensity to comment. In conjunction, these traits can cause the recommendation algorithm to “misinterpret” a user’s negative comment as a successful recommendation for that viewer. This “misinterpretation” can create a feedback loop, where the recommendation algorithm will show the content to other hostile users with similar “anti-preferences.” Expectedly, this hostile audience would share a similar high propensity to leave hostile comments of their own, thus restarting the loop. From the content creator’s side, the anti-preference phenomenon can appear as a sudden and seemingly systematic increase in hostile comments, similar to the experiences described within the “reaching the wrong audience” folk theory.

AI in semi-automated journalism: a review of Indonesia's journalistic ethics

In the past five years, Artificial Intelligence (AI) has entered Indonesian newsrooms, aligning with the rise of Industry 4.0. This integration is poised to redefine the roles of journalists and media professionals in the future. Semi-automated journalism, a blend of human journalistic practices and AI technology, is being used in newsrooms to create news content. However, it faces challenges such as accuracy, ethics, and concerns about bias. This article uses a comprehensive literature review, covering academic publications and industry reports from the past five years, to examine semi-automated journalism in the AI era, with a focus on journalistic ethics. The review includes sources selected based on relevance, credibility, and contribution to the topic. The findings identify specific challenges such as the need for transparency in AI processes, potential biases in algorithmic decision-making, and the impact on journalistic integrity. While AI can improve efficiency, it falls short in content quality. It is crucial to train journalists for responsible AI use. Ethical principles should guide AI use in journalism to ensure the provision of quality and valuable information.

Procedural fairness in algorithmic decision-making: the role of public engagement

Despite the widespread use of automated decision-making (ADM) systems, they are often developed without involving the public or those directly affected, leading to concerns about systematic biases that may perpetuate structural injustices. Existing formal fairness approaches primarily focus on statistical outcomes across demographic groups or individual fairness, yet these methods reveal ambiguities and limitations in addressing fairness comprehensively. This paper argues for a holistic approach to algorithmic fairness that integrates procedural fairness, considering both decision-making processes and their outcomes. Procedural fairness emphasizes the importance of fair decision-making procedures, which aligns with theories of relational justice that stress the quality of social relations and power dynamics. We highlight the need for substantive procedural fairness to ensure better outcomes and address forward-looking responsibilities. Additionally, we propose leveraging Public Engagement, a core dimension within the well-established Responsible Research and Innovation framework, to enhance procedural fairness in ADM systems. Our contribution underscores the value of Public Engagement in fostering fairer ADM processes, thereby expanding the current focus beyond technical outcome-based approaches to encompass broader procedural considerations.

Designing a machine learning-based lending model to enhance access to capital for small and medium enterprises

This paper presents the design of a machine learning-based lending model aimed at enhancing access to capital for Small and Medium Enterprises (SMEs). SMEs play a crucial role in driving economic growth, yet they often face significant barriers in securing funding due to traditional credit assessment methods that rely heavily on financial history and collateral. By leveraging machine learning (ML) algorithms, this proposed model incorporates diverse data sources, including financial statements, market trends, customer behavior, and non-traditional data such as social media activity and payment histories, to better assess creditworthiness. The model utilizes supervised learning techniques, such as decision trees, random forests, and neural networks, to analyze this diverse data and identify patterns that traditional models may overlook. Additionally, it integrates real-time data processing to continuously update credit profiles, allowing for more dynamic and responsive lending decisions. The primary aim is to improve the accuracy of risk assessment, reduce default rates, and ultimately facilitate more inclusive lending practices for SMEs. One of the core advantages of the ML-based model is its ability to offer personalized lending terms and products by identifying the unique characteristics and risk factors of each SME. This approach not only improves access to capital for underbanked businesses but also ensures that lenders can manage risk more effectively. The paper also discusses the potential ethical considerations, such as ensuring data privacy and avoiding biases in algorithmic decision-making, which are critical for the responsible implementation of such a system. In conclusion, this machine learning-based lending model presents an innovative solution to the long-standing challenges SMEs face in accessing capital. By harnessing the power of data analytics and advanced algorithms, it paves the way for more equitable and efficient lending processes, supporting the growth and sustainability of SMEs. Keywords: Machine Learning, Lending Model, Access To Capital, Small And Medium Enterprises, Risk Assessment, Creditworthiness, Financial Inclusion, Data Privacy, Algorithmic Decision-Making, Supervised Learning.

THE NEED FOR LEGAL REGULATION OF THE ‟BLACK BOX” OF ARTIFICIAL INTELLIGENCE

This paper discusses the transparency and regulation of Artificial Intelligence (hereinafter the AI). It discusses the growing trend of using decision-making algorithms in various areas of life and the need for transparent and the understandable AI systems. The author highlights the concept of a ‟black box” where algorithms and their processes become hidden to users and developers, and emphasises the importance of explainability and accessibility of information about the operation of the AI. The article also addresses the right to explanation and possible strategies for ensuring transparency and accountability in algorithmic decision-making, including parameter documentation and system certification.

Definitions of Fairness Differ Across Socioeconomic Groups & Shape Perceptions of Algorithmic Decisions

Understanding how people perceive algorithmic decision-making is remains critical, as these systems are increasingly integrated into areas such as education, healthcare, and criminal justice. These perceptions can shape trust in, compliance with, and the perceived legitimacy of automated systems. Focusing on San Francisco's decade-long policy of algorithmic school assignments, we draw on procedural and distributive justice theory to investigate parents' fairness perceptions of San Francisco's school assignment system. We find that key differences in parents' definitions of fairness shape their preferences for what constitutes a fair school assignment system, while also controlling for parents' school assignment outcomes. Moreover, parents' definitions and perceptions of fairness differ across socioeconomic and racial groups. For instance, among white respondents, the most used definition of fairness was "proximity'' to their assigned school, whereas, among Hispanic or Latino parents, the most popular definition of fairness was that the "same rules'' are applied to everyone. It is crucial for computational system designers and policymakers to consider these differences when deciding on the goals and values embedded in decision-making systems and who those goals and values reflect.

Insights into Algorithmic Decision-Making Systems via a Decolonial-Intersectional Lens: A Cross-Analysis Case Study

As the line between governance, technology, and decision-making becomes ever-blurred, advocates and disparagers of Artificial Intelligence (AI) debate the impacts of its societal integration. Despite efforts to promote Fairness, Accountability, Transparency, and Ethics (FATE), vulnerable populations continue to be systematically marginalized and made “invisible” by the racialised, classed, and colonial institutions buttressing Algorithmic Decision-Making Systems (ADMS). To address these sociotechnical risks and acknowledge our privileged, Western “standpoint epistemology,” we employ a “metaparadigm perspective” to engage the literature and case studies through a critical theory lens. The cross-analysis of three case studies: Systems Theory, the United States’ “Blueprint for an AI Bill of Rights,” and predictive policing demonstrate that our current risk mitigation frameworks are incapable of addressing transhistorical injustices. Therefore, we call for operationalizing intersectional risk theory via the decolonial turn to address ADMS risks. On that note, we propose three algorithmic accountability recommendations. We theorize that through a reformulation of FATE principles to that of the algorithmic (decolonial) self, understanding power imbalances through marginalized risk discourses, and advancing intersectional ML decision-making practices, policymakers may be better positioned to enact equitable and just ADMS.

Navigating ethical challenges of explainable ai in autonomous systems

The rapid integration of autonomous systems, such as vehicles, drones, and robots, into various sectors brings forth significant ethical challenges concerning their decision-making processes. This paper examines the role of Explainable AI (XAI) in addressing these challenges, particularly regarding accountability in the event of accidents and the necessity of human oversight in automated environments. We discuss the critical ethical implications of transparency, emphasizing how XAI can bridge the gap between complex algorithmic decision-making and public understanding, thereby fostering trust in these technologies. The paper also outlines the current regulatory frameworks for AI safety, analysing their effectiveness in promoting responsible innovation. Furthermore, we investigate the consequences of opaque algorithms, especially in life-critical applications where the stakes are exceptionally high. Through an analysis of case studies, we showcase how organizations have successfully implemented XAI to enhance safety measures and uphold ethical responsibility in their autonomous systems. Ultimately, this study advocates for the integration of XAI as a vital component in developing responsible autonomous technologies, ensuring accountability, and safeguarding public trust in an era increasingly defined by automation.

Knowledge and support for AI in the public sector: a deliberative poll experiment

AbstractWe are on the verge of a revolution in public sector decision-making processes, where computers will take over many of the governance tasks previously assigned to human bureaucrats. Governance decisions based on algorithmic information processing are increasing in numbers and scope, contributing to decisions that impact the lives of individual citizens. While significant attention in the recent few years has been devoted to normative discussions on fairness, accountability, and transparency related to algorithmic decision-making based on artificial intelligence, less is known about citizens’ considered views on this issue. To put society in-the-loop, a Deliberative Poll was thus carried out on the topic of using artificial intelligence in the public sector, as a form of in-depth public consultation. The three use cases that were selected for deliberation were refugee reallocation, a welfare-to-work program, and parole. A key finding was that after having acquired more knowledge about the concrete use cases, participants were overall more supportive of using artificial intelligence in the decision processes. The event was set up with a pretest/post-test control group experimental design, and as such, the results offer experimental evidence to extant observational studies showing positive associations between knowledge and support for using artificial intelligence.

Sensitive personal health information, public trust, and biases in algorithmic decision-making: public health concerns in digital wellness.

Sensitive personal health information, public trust, and biases in algorithmic decision-making: public health concerns in digital wellness.

Algorithmic Decision-Making under Agents with Persistent Improvement

This paper studies algorithmic decision-making under human strategic behavior, where a decision-maker uses an algorithm to make decisions about human agents, and the latter with information about the algorithm may exert effort strategically and improve to receive favorable decisions. Unlike prior works that assume agents benefit from their efforts immediately, we consider realistic scenarios where the impacts of these efforts are persistent and agents benefit from efforts by making improvements gradually. We first develop a dynamic model to characterize persistent improvements and based on this construct a Stackelberg game to model the interplay between agents and the decision-maker. We analytically characterize the equilibrium strategies and identify conditions under which agents have incentives to invest efforts to improve their qualifications. With the dynamics, we then study how the decision-maker can design an optimal policy to incentivize the largest improvements inside the agent population. We also extend the model to settings where 1) agents may be dishonest and game the algorithm into making favorable but erroneous decisions; 2) honest efforts are forgettable and not sufficient to guarantee persistent improvements. With the extended models, we further examine conditions under which agents prefer honest efforts over dishonest behavior and the impacts of forgettable efforts.

Non-linear Welfare-Aware Strategic Learning

This paper studies algorithmic decision-making in the presence of strategic individual behaviors, where an ML model is used to make decisions about human agents and the latter can adapt their behavior strategically to improve their future data. Existing results on strategic learning have largely focused on the linear setting where agents with linear labeling functions best respond to a (noisy) linear decision policy. Instead, this work focuses on general non-linear settings where agents respond to the decision policy with only "local information" of the policy. Moreover, we simultaneously consider objectives of maximizing decision-maker welfare (model prediction accuracy), social welfare (agent improvement caused by strategic behaviors), and agent welfare (the extent that ML underestimates the agents). We first generalize the agent best response model in previous works to the non-linear setting and then investigate the compatibility of welfare objectives. We show the three welfare can attain the optimum simultaneously only under restrictive conditions which are challenging to achieve in non-linear settings. The theoretical results imply that existing works solely maximizing the welfare of a subset of parties usually diminish the welfare of others. We thus claim the necessity of balancing the welfare of each party in non-linear settings and propose an irreducible optimization algorithm suitable for general strategic learning. Experiments on synthetic and real data validate the proposed algorithm.

Why Do Swiss HR Departments Dislike Algorithms in Their Recruitment Process? An Empirical Analysis

This study investigates the factors influencing the aversion of Swiss HRM departments to algorithmic decision-making in the hiring process. Based on a survey provided to 324 private and public HR professionals, it explores how privacy concerns, general attitude toward AI, perceived threat, personal development concerns, and personal well-being concerns, as well as control variables such as gender, age, time with organization, and hierarchical position, influence their algorithmic aversion. Its aim is to understand the algorithmic aversion of HR employees in the private and public sectors. The following article is based on three PLS-SEM structural equation models. Its main findings are that privacy concerns are generally important in explaining aversion to algorithmic decision-making in the hiring process, especially in the private sector. Positive and negative general attitudes toward AI are also very important, especially in the public sector. Perceived threat also has a positive impact on algorithmic aversion among private and public sector respondents. While personal development concerns explain algorithmic aversion in general, they are most important for public actors. Finally, personal well-being concerns explain algorithmic aversion in both the private and public sectors, but more so in the latter, while our control variables were never statistically significant. This said, this article makes a significant contribution to explaining the causes of the aversion of HR departments to recruitment decision-making algorithms. This can enable practitioners to anticipate these various points in order to minimize the reluctance of HR professionals when considering the implementation of this type of tool.

Inevitable challenges of autonomy: ethical concerns in personalized algorithmic decision-making

Breakthroughs in Artificial Intelligence (AI) have transformed human decision-making processes. Individuals increasingly rely on algorithms to assist with various tasks such as searching and shopping. However, the widespread use of personalized algorithmic decision-making has raised numerous ethical concerns, specifically its impact on user autonomy. This article examines these concerns and argues that algorithmic decision-making presents several challenges to user autonomy that are difficult to eliminate. These challenges include the fact that algorithms deviate from a user’s authentic self, create self-reinforcing loops that narrow the user’s self, and lead to a decline in the user’s capacities. This article attributes these challenges to the ontological differences between users and AI, as well as their unawareness regarding the ethical risks associated with algorithms, asserting that the sense of autonomy users experience in algorithmic decision-making is merely an illusion. From the perspective of human-AI interaction, it is proposed a strategy that achieves a balance between human insight and AI through a human-in-the-loop approach to mitigate these challenges. This approach maximizes the strengths of AI in processing and analyzing data while ensuring that users retain agency.