Tristan Lim

Financial-services AI governance often relies on disclosing artificial intelligence (AI) involvement, yet it is unclear whether customers identify who made the decision. In a 3 × 2 vignette experiment on a partially approved insurance claim, 402 respondents were randomly assigned to a disclosed decision authority (officer decides on an AI recommendation; AI decides with brief officer review; AI decides alone) and to a lower-or higher-impact scenario. Only 39.8% chose the answer matching the disclosure. "AI with human review" was the most common answer in every condition. Under condition-invariant default responses, a post hoc threshold-with-guessing model estimated a disclosure-consistent component of .14 (95% CI .07-.21). Partial identification shows that its size, and even its positivity, depends on how far defaults differ across conditions. Disclosed authority had no detectable average effect on procedural fairness. The higher-impact scenario package lowered fairness. As hypothesised, lower fairness was associated with higher escalation preference. The study separates disclosed from attributed decision authority. Under stated assumptions, it shows how far manipulationcheck match rates can reflect default responding. Insurers and regulators may want to test whether customers identify AI authority after disclosure.

Multi-criteria decision-making methods conventionally rank alternatives by aggregating normalized criterion values into a composite score, an approach that depends on the normalization scheme, requires externally determined criterion weights, and evaluates performance in absolute terms without accounting for inter-criteria dependencies. This paper proposes machine learning-based Peer-Prediction Trees for multi-criteria decision-making, a novel method that ranks alternatives by their excess performance over data-driven peer expectations. For each criterion, a leave-one-out cross-validated decision tree predicts each alternative’s value from its performance on all other criteria; the residual between observed and predicted values extracts how much the alternative outperforms the empirical trade-off structure constraining its peers. Residuals are standardized, aligned with the preferred criterion direction, and aggregated into an Excess Performance Index using uniqueness-based weights derived endogenously from the cross-validated predictive fit. The method is validated on three real-world datasets: a numeric vehicle selection problem (22 alternatives, five criteria), a fully linguistic financial performance evaluation of Turkish listed firms (32 alternatives, four criteria), and a mixed numeric-linguistic exchange-traded fund selection problem (12 alternatives, nine criteria). The contributions of this work are fourfold: (1) the introduction of peer prediction via leave-one-out cross-validated decision trees as a multi-criteria decision-making ranking paradigm that evaluates excess performance rather than absolute scores; (2) a uniqueness-based criterion weighting scheme embedded into the method, eliminating the need for separate weighting procedures; (3) native handling of numeric, linguistic, and mixed decision matrices within a single unified framework; and (4) empirical validation across three heterogeneous real-world applications demonstrating the method’s versatility and interpretability.

Consumer financial services depend on interconnected institutions, products and regulatory processes, yet complaints about failures within this infrastructure are usually analysed as individual cases or grouped using administrative categories. This study examines whether large-scale complaint data can instead reveal recurring structures of failure. Using 4.7 million complaints submitted to the U.S. Consumer Financial Protection Bureau between 2011 and 2024, we develop a Failure Ecosystem Framework that groups complaint narratives by semantic similarity and traces the institutions and products associated with each group over time. The analysis identifies 16 failure ecosystems, spanning credit reporting, identity theft, debt collection, payment disputes, mortgage servicing and digital finance. These ecosystems differ substantially: some failures concentrate among a few institutions while others are widely distributed, and in many, the same institution–product combinations recur from year to year. Greater concentration is not systematically associated with better or worse complaint outcomes, and the evidence is least conclusive for monetary compensation. Corrective outcomes vary more across failure types than with concentration, indicating that market structure alone cannot explain how consumer failures are remedied. Complaints can therefore reveal persistent structures of failure that case-by-case review does not expose, supporting system-level monitoring of vulnerabilities in consumer finance.

Artificial intelligence (AI) is increasingly recognised for its transformative implications for sustainability transitions. Yet little is known about how AI co-evolves with green technological innovation systems and whether institutional adaptation keeps pace with technological diffusion. This study maps 3357 peer-reviewed publications between 2003 and 2025 using transformer-based topic modelling and cross-model triangulation to characterise structural evolution across enabling technologies, sectoral applications, and governance domains. Results reveal a reproducible triadic configuration consistent with innovation-system layering, alongside pronounced asymmetry in growth trajectories. While AI-enabled application domains (e.g., energy systems, waste and circularity, agriculture, and urban mobility) exhibit sustained expansion and increasing specialisation, governance and institutional strands demonstrate thinner density, greater model sensitivity, and delayed acceleration. These patterns are consistent with an asynchronous relationship between technological capability and institutional oversight, echoing the innovation-regulation lag observed in other technology-diffusion processes. The findings contribute to technological change literature by providing systematic evidence on the thematic structure and evolution of AI-enabled sustainability research and by proposing a co-evolution diagnostic framework for monitoring alignment between technological expansion and governance attention.

Artificial intelligence (AI) is increasingly embedded within financial decision making, yet existing research has largely focused on adoption, trust, reliance, and performance. Consequently, limited attention has been devoted to how AI systems influence the formation of financial judgments. This study develops Delegated Financial Cognition (DFC), defined as the extent to which individuals allocate epistemic authority to algorithmic agents when forming financial judgments. Drawing on a PRISMA-guided systematic review and conceptual synthesis across five literature streams, the study develops a process theory explaining how evaluative authority is allocated, reinforced, and moderated within AI-mediated financial environments. The framework contributes to behavioral finance by identifying authority allocation as a missing mechanism linking AI capability, trust, reliance, and longterm financial capability.

Multi-criteria decision-making (MCDM) methods conventionally rank alternatives by aggregating normalized criterion values into a composite score, an approach that depends on the normalization scheme, requires externally determined criterion weights, and evaluates performance in absolute terms without accounting for inter-criteria dependencies. This paper proposes Peer-Prediction Trees (PPT) for MCDM, a novel method that ranks alternatives by their excess performance over data-driven peer expectations. For each criterion, a leave-one-out cross-validated (LOOCV) decision tree predicts each alternative's value from its performance on all other criteria; the residual between observed and predicted values measures how much the alternative outperforms the empirical trade-off structure constraining its peers. Residuals are standardized, aligned with the preferred criterion direction, and aggregated into an Excess Performance Index (EPI) using uniqueness-based weights derived endogenously from the cross-validated predictive fit. The method is validated on three real-world datasets: a numeric vehicle selection problem (22 alternatives, 5 criteria), a fully linguistic financial performance evaluation of Turkish listed firms (32 alternatives, 4 criteria), and a mixed numeric-linguistic exchange-traded fund (ETF) selection problem (12 alternatives, 9 criteria). The contributions of this work are four-fold: (1) the introduction of peer prediction via LOOCV decision trees as an MCDM ranking paradigm that evaluates excess performance rather than absolute scores; (2) a uniqueness-based criterion weighting scheme embedded into the method, eliminating the need for separate weighting procedures; (3) native handling of numeric, linguistic, and mixed decision matrices within a single unified framework; (4) empirical validation across three heterogeneous real-world applications demonstrating the method's versatility and interpretability.

The increasing integration of agentic artificial intelligence into digital financial platforms, such as robo-advisors, AI-driven credit systems, and conversational financial agents, is transforming how financial decisions are made and executed. These systems extend beyond traditional decision support by exhibiting goal-directed behavior, adaptive learning, and autonomous execution, functioning as quasi-independent decision agents. As a result, users increasingly delegate evaluative financial reasoning to algorithmic systems, shifting the locus of cognition from humans to algorithms. This paper develops a process theory of Delegated Financial Cognition (DFC) to explain how perceived algorithmic agency induces cognitive delegation, reduces monitoring, and shapes trust calibration. The framework theorizes how miscalibrated trust interacts with behavioral biases, leading to amplification or attenuation in financial outcomes. At scale, these dynamics may induce coordinated behavior, reduce decision diversity, and amplify volatility. The study proposes testable propositions and specifies boundary conditions across technological, human, and environmental dimensions. By integrating behavioral finance and trust in automation, it contributes a process-based explanation of human-AI decision-making in algorithmic financial contexts.

This study proposes a novel data-driven machine learning (ML) framework for multi-criteria environmental, social, and governance (ESG) evaluation. The framework aims to address the transparency, consistency, and subjectivity limitations of existing ESG evaluation systems by employing a fully data-driven, modular, and ML-supported architecture. It comprises three main modules: (i) ESG data preprocessing with missing-data imputation by the MissForest algorithm; (ii) a three-plane ESG feature selection workflow that integrates clustering, feature importance, and classification algorithms to identify representative ESG indicators; and (iii) a hybrid weighting and ranking procedure that combines unsupervised principal component analysis (PCA), criteria importance through inter-criteria correlation (CRITIC), and technique for order preference by similarity to ideal solution (TOPSIS) methods. A recent 2024 real-world application involving 57 listed Chinese pharmaceutical and biotechnology companies and 70 ESG indicators demonstrates the framework’s practical utility in producing transparent and objective ESG rankings. The main contributions of this work are fourfold: (1) the development of an end-to-end, entirely data-driven ML framework for ESG evaluation; (2) the introduction of an innovative three-plane ESG feature selection workflow within the framework; (3) the first study for designing a hybrid PCA-CRITIC-TOPSIS approach in ESG weighting and ranking; (4) the validation of the framework through a real-world industry application using recent and authentic ESG data.

The increasing integration of agentic artificial intelligence into digital financial platforms and electronic markets, such as robo-advisors, AI-driven credit systems, and conversational financial agents, is transforming how financial decisions are made and executed. These systems extend beyond traditional decision support by exhibiting goal-directed behavior, adaptive learning, and autonomous execution, functioning as quasi-independent decision agents within digital financial ecosystems. As a result, users increasingly delegate evaluative financial reasoning to algorithmic systems, shifting the locus of cognition from human decision-makers to platform-embedded intelligence. This paper develops a process theory of Delegated Financial Cognition (DFC) to explain how perceived algorithmic agency induces cognitive delegation, reduces user monitoring, and shapes trust calibration in agentic fintech market environments. The framework further theorizes how miscalibrated trust interacts with behavioral biases, leading to amplification or attenuation effects in financial decision outcomes. The study formalizes possible research propositions and identify boundary conditions across technological (e.g., transparency, explainability), human (e.g., financial literacy, algorithm familiarity), and environmental (e.g., market volatility, decision complexity) dimensions. By integrating behavioral finance, trust in automation, and information systems research, the study aims to contribute an explanation of how human-AI decision processes occur on agentic fintech platforms. The findings look to advance understanding of digital financial services by conceptualizing cognitive delegation as a core mechanism underlying emerging forms of embedded and contextual financial decision-making.