Privacy Risks Research Articles

Continual Learning From a Stream of APIs.

Continual learning (CL) aims to learn new tasks without forgetting previous tasks. However, existing CL methods require a large amount of raw data, which is often unavailable due to copyright considerations and privacy risks. Instead, stakeholders usually release pre-trained machine learning models as a service (MLaaS), which users can access via APIs. This paper considers two practical-yet-novel CL settings: data-efficient CL (DECL-APIs) and data-free CL (DFCL-APIs), which achieve CL from a stream of APIs with partial or no raw data. Performing CL under these two new settings faces several challenges: unavailable full raw data, unknown model parameters, heterogeneous models of arbitrary architecture and scale, and catastrophic forgetting of previous APIs. To overcome these issues, we propose a novel data-free cooperative continual distillation learning framework that distills knowledge from a stream of APIs into a CL model by generating pseudo data, just by querying APIs. Specifically, our framework includes two cooperative generators and one CL model, forming their training as an adversarial game. We first use the CL model and the current API as fixed discriminators to train generators via a derivative-free method. Generators adversarially generate hard and diverse synthetic data to maximize the response gap between the CL model and the API. Next, we train the CL model by minimizing the gap between the responses of the CL model and the black-box API on synthetic data, to transfer the API's knowledge to the CL model. Furthermore, we propose a new regularization term based on network similarity to prevent catastrophic forgetting of previous APIs. Our method performs comparably to classic CL with full raw data on the MNIST and SVHN datasets in the DFCL-APIs setting. In the DECL-APIs setting, our method achieves 0.97×, 0.75× and 0.69× performance of classic CL on the more challenging CIFAR10, CIFAR100, and MiniImageNet, respectively.

Cascade Vertical Federated Learning Towards Straggler Mitigation and Label Privacy Over Distributed Labels

Vertical federated learning (VFL) enables collaborative machine learning on vertically partitioned data with privacy-preservation. Most VFL methods face three daunting challenges in real-world applications. First, most existing VFL methods assume that at least one party holds the complete set of labels of all data samples. However, this assumption often violates the nature of many practical scenarios, where the parties only have partial labels. Second, the heterogeneity and dynamic of computational and communication resources in participated parties may cause the straggler problem and slow down training convergence. Third, the confidential label information could be exposed through malicious parties during VFL. To address these challenges, we propose a novel VFL algorithm named Cascade Vertical Federated Learning (CVFL), in which partitioned labels can be fully utilized to train neural networks with privacy-preservation. To mitigate the straggler problem, we design a novel optimization objective to increase straggler's contribution to the trained models. To mitigate the label privacy risks, we design a novel defense approach to protect the label privacy of CVFL. We conduct comprehensive experiments and the results demonstrate the effectiveness and efficiency of CVFL. Further, the proposed defense approach can achieve a better tradeoff between label privacy and model utility than two widely-used defense approaches.

Assessing the role of artificial intelligence in enhancing customer personalization: A study of ethical and privacy implications in digital marketing

Technological evolution, especially in artificial intelligence (AI), has significantly changed digital marketing, specifically customer-targeted experience. Machine learning, recommendation systems, and predictive analytics have made it possible for enterprises to personalize content, products, and services, among other things, on a very large scale. Nevertheless, all these advances have come with large ethical and privacy concerns that are yet to be met. This paper discusses how AI can enhance customers' personalization to analyze the moral and privacy considerations involved. The research examines how machine learning marketing strategies work in the digital environment and real-life examples of organizations that successfully implement these approaches and potential issues. It highlights how individuals might buy more products and make businesses more profitable with customized advertising techniques. However, the paper also describes the moral problems of AI that use personal data without permission, artificially intelligent algorithms that affect decision-making, and artificial intelligence decision-making. At the same time, its work has yet to be well known. The paper looks into privacy risks like data breach incidents, privacy violations, GDPR, and CCPA. This paper adopts an exploratory case study research design as it gathers academic papers, industry examples, and expert opinions to present a balanced view of the effects of AI. Based on the research, conclusions can be drawn that great opportunities AI opens for personalization should be met with safety considerations and working ethical guidelines. In the end, the paper presents guidelines that should help businesses integrate trust and compliance into creating and using AI technologies.

Evaluating the benefits and risks of facial recognition payment systems adoption in Malaysia

PurposeCurrent developments in the FinTech payment industry have shown a rapid revolution in Industry 4.0, and understanding the factors affecting individual acceptance of facial recognition payment (FRP) is crucial. Hence, this study aims to evaluate the benefits and risks of FRP system adoption in Malaysia.Design/methodology/approachThe perceived risks and benefits framework is adopted as the foundation in this study to examine the various risks and benefits that users perceive, along with the trust factor, to study the relationships between these variables. Data were collected via an online questionnaire, and the hypotheses were tested using Partial Least Squares analysis on 277 responses.FindingsThe results revealed that perceived risk is a significant predictor of users' intention to use the FRP system. Privacy risk and financial risk significantly influence perceived risks, while security risk does not. Although convenience, perceived ease of use and perceived trust positively influence perceived benefits, perceived benefits do not significantly influence adoption intention. Moreover, perceived trust negatively affects perceived risks while positively affecting both perceived benefits and adoption intention. Additionally, personal innovativeness moderates the relationship between perceived risks and the intention to use the FRP system.Practical implicationsThis study helps policymakers and service providers understand individuals’ concerns and expectations regarding FRP systems. It aids practitioners in developing strategies to build trust, address innovativeness differences and mitigate risks, serving as a roadmap for integrating these systems into Malaysia's financial landscape.Originality/valueThis study distinguishes itself from prior research by evaluating FRP system adoption in Malaysia through the lens of perceived risks and benefits framework. It also explores personal innovativeness as a moderator, examining its impact on the relationship between usage intention and perceived risks and benefits. Additionally, it highlights perceived trust as a crucial factor influencing individuals' intention to adopt FRPs.

Do the technological anxiety, privacy and physical risks matter in retail customers'‏ ‏adoption of AR‏ ‏apps? An extended UTAUT2 approach

PurposeDue to augmented reality (AR) technology improvements, the retail industry has embraced smart retailing as its primary business model. Therefore, organizations must comprehend the intricacies of AR adoption to persuade clients to adopt this revolutionary technology effectively. Thus, the current study proposes and evaluates a comprehensive model that includes unified theory of acceptance and use of technology (UTAUT2), privacy concerns, physical risks and technological anxiety to predict customers’ intention to use AR apps in the retail industry in the Egyptian context.Design/methodology/approachThe current study examines 398 responses from Egyptian shoppers using partial least squares structural equation modeling (PLS-SEM). Snowball sampling was employed in the existing study. The participants were selected using a “self-selection” strategy, which is an excellent method for research investigations in which the participants freely participate.FindingsConsumers’ intentions to use AR apps in retail settings are positively impacted by task-technology fit, performance expectation, effort expectancy, social influence, facilitating conditions and hedonic motivation. Conversely, privacy and physical risks negatively affect customers’ intention to use AR apps in retail. Furthermore, technological anxiety serves as a moderator factor in these connections.Originality/valueTo the best of our knowledge, the current study is considered the first to test the role of UTAUT2, privacy and physical risks on users' behavioral intentions toward adopting AR apps in retail. It also examines technological anxiety as a moderator in the retail setting.

FHG-PR: a hybridized fuzzy-AHP and game theory model for assessing privacy risk on social media platforms

Social media platforms have become integral to our daily lives, enabling global connectivity and interaction. Current research provides general solutions for mitigating privacy risks, but it does not focus on particular platforms. However, the voluntary sharing of personal information on these platforms presents considerable privacy risks, highlighting the need for effective risk assessment mechanisms. This study presents a hybrid approach that uses the fuzzy Analytical Hierarchy Process (AHP) and game theory to estimate privacy risks on Meta and X. We used a fuzzy AHP to rank determinant variables based on surveys of social media users and professionals. These weighted criteria were then used in a Multi-Criteria Decision-Making (MCDM) framework using cooperative game theory to discover ways to lower privacy hazards. The model accurately analyzed the privacy threats on Meta and X, recommending alternate techniques for preventing breaches in data privacy. The cooperative game theory method allows stakeholders to collaborate on creating privacy-preserving measures. The findings emphasize the model’s ability to address platform-specific privacy threats, as well as its flexibility in other social media settings. The model’s potential for real-world application is demonstrated by its ability to provide practical risk reduction measures, which improve privacy protection for social media users.

The Impact of Consumer Perceptions of Privacy and Security Risks on Re-purchasing Intention for Online Shopping in Chiang Mai Province, Thailand

Background and Aim: Chiang Mai Province, a growing economic center in northern Thailand, is a preferred mode of consumerism due to its convenience and wide product selection, but faces barriers to adoption due to concerns about these risks. This research investigates the impact of consumer perceptions of privacy and security risks on re-purchasing intention for online shopping in Chiang Mai Province, Thailand. The study aims to understand how these perceptions influence consumer behavior in the digital marketplace. Materials and Methods: A quantitative cross-sectional survey design was employed, with data collected from 385 online shoppers in Chiang Mai Province. The research instrument consisted of a structured online questionnaire assessing perceived privacy and security risks, as well as re-purchasing intentions. Multiple regression analysis was used to test the hypotheses. Results: The findings reveal that both perceived privacy and security risks significantly influence re-purchasing intention. Consumers who perceive higher levels of privacy and security are more likely to engage in repeat online purchases. The model explained a substantial proportion of the variance in re-purchasing intention, highlighting the importance of these factors in shaping online shopping behavior. Conclusion: This research provides valuable insights for e-commerce businesses and policymakers in Chiang Mai Province. By prioritizing and investing in measures that enhance consumer perceptions of privacy and security, businesses can foster trust, encourage repeat purchases, and drive sustainable growth in the online marketplace.

Smart Shopping Carts in Food Retailing: Innovative Technology and Shopping Experience in Stationary Retail

ABSTRACTIn the development of the Internet of Things, smart shopping carts present a digital innovation that can realize cost‐cutting potential and improve the shopping experience. The impact of smart shopping carts on customers' shopping experience is analyzed in stationary grocery retailing. We explore first whether customers are willing to use these technological innovations, specifically smart shopping carts. Second, the effect of smart shopping carts on customers' shopping experiences is evaluated. For this, we employed a mixed method approach combining a survey study and an interview follow‐up study. A purposive sampling recruited 291 stationary grocery shoppers. A variance‐based structural equation analysis tests the research model for the adoption of smart shopping carts and their relevance to customers' shopping experience. Next, 22 interviews were conducted with grocery shoppers following an intervention approach. Customers are willing to use smart shopping carts when shopping for groceries in stationary retailing. Customers focus on the ease of use and usefulness of the shopping carts, whereas privacy risk and temporary enjoyment are less relevant. However, our empirical results show that affective, social (employee), and physical components predominantly shape the shopping experience compared with the technical component of smart shopping carts. Smart shopping carts similarly affect the cognitive component and the social (customer) component. In contrast to other digital retail innovations, customers seem to value utility of smart shopping carts above else. Stationary retailers can, thus, focus their introduction of smart shopping carts on countering technology anxiety and highlighting customer benefits of shopping comfort and reduced checkout times.

Financial Inclusion and Climate Resilience: The Role for an AI-Enhanced Digital Wallet in Caribbean SIDS

Caribbean Small Island Developing States (SIDS) are highly vulnerable to extreme weather events and climate change. Caribbean SIDS climate vulnerability is worsened by their high level of financial exclusion. Many people do not have bank accounts and access to electronic fund transfer (EFT). As such, they cannot electronically receive funds before or after a natural disaster to cope with the effects. The financial exclusion problem can be addressed through a digital wallet. A digital wallet is a financial transaction application that securely stores a user’s banking and payment information on a cloud interface and allows the user to perform a transaction while hiding their banking information from a vendor. The biggest concern of users with regards to the use of digital wallets are its convenience and security. While digital wallets offer outstanding convenience of purchasing goods and services, data privacy and fraud risks deter people from adopting mobile payment. Potential fraud risk to digital wallets can be identified with anomaly detection techniques. The research problem investigated in this study is how the implementation of an artificial intelligence (AI) enhanced digital wallet can facilitate financial inclusion in the Caribbean, particularly in the context of disaster preparedness and recovery. Since security is an important aspect of a digital wallet, a sub-objective of this study is to derive an appropriate model for anomaly detection in financial transactions in a digital wallet. This study modifies Liu et al. [1] Gated Transformer Networks (GTN) architecture to allow for univariate time series classification. The corresponding new model is referred to as the Univariate Gated Transformer Network (UGTN). The UGTN is used for anomaly detection in financial data from a digital wallet. This study also provides policy recommendations for the implementation of a digital wallet to facilitate financial inclusion and climate resilience in Caribbean SIDS.

A data privacy survey on personal identifiable information (PII) left on rental vehicle infotainment systems

In the current automotive ecosystem, the trend of pairing mobile devices to connected vehicles is gaining momentum, providing a vast number of benefits such as hands-free driving and remote vehicle control. However, along with these conveniences arises the issue of data accumulation, ranging from vehicle diagnostics to personal identifiable information (PII). The problem emerges when a consumer rents a vehicle, pairs their mobile device to the infotainment system, and neglects to remove their device prior to returning the vehicle. This oversight can potentially expose vulnerabilities with the current renter’s PII for subsequent renters to exploit. Research indicates that renters often overlook the deletion process prior to returning the rental vehicle and are unaware of whose responsibility it is to perform this task. In this survey, we investigated the experiences and perceptions of a group of consumers who have previously rented vehicles. We wanted to know if the participants were aware they were responsible for deleting their mobile device from the rental vehicle before returning it, the renters’ importance and tolerance for risk if they overlooked the deletion process leaving their shared data on the infotainment system for subsequent users, and if they were aware who was responsible for deleting their shared data. Lastly, we explored if the participants supported an automated solution to perform this manual deletion process. The goal of this paper is to demonstrate the significance of safeguarding PII in connected vehicles and advocate for the adoption of an automated solution to mitigate this privacy risk.

Five Dimensions of AI Readiness (AIR-5D) Framework- A Preparedness Assessment Tool for Healthcare Organizations

Background: Artificial Intelligence (AI) has transformative potential in healthcare, and it is very useful in areas such as drug discovery, diagnostics, and patient management. However, there is a lack of tools to assess healthcare organizations’ readiness to adopt AI technologies. This study introduces the AI Readiness Five Dimension (AIR-5D) framework, addressing this gap. Methods: The AIR-5D framework was developed using a two-step process: identifying dimensions of AI readiness from literature and weighing these dimensions through expert focus groups. The Analytical Hierarchy Process (AHP) was employed to calculate weights, ensuring consistency and reliability. Results: The results identified five key dimensions: Opportunity Discovery (0.44), Data Management 90.22), IT Environment and Security (0.194), Risk Privacy and Governance (0.101), and Adoption of Technology (0.043). “Opportunity Discovery” was the most critical dimension, while “Adoption of Technology” ranked lowest. Six case studies demonstrated varying AI readiness (score between 3 and 4 on a scale of 5), highlighting challenges in moving beyond AI collaboration to optimization. Conclusions: The AIR-5D framework offers a structured approach for healthcare organizations to assess and enhance their AI readiness. It emphasizes the importance of understanding value, robust data management, and strategic alignment in successful AI adoption.

Advice from a Doctor or AI? Understanding Willingness to Disclose Information Through Remote Patient Monitoring to Receive Health Advice

Remote Patient Monitoring (RPM) devices transmit patients' medical indicators (e.g., blood pressure) from the patient's home testing equipment to their healthcare providers, in order to monitor chronic conditions such as hypertension. AI systems have the potential to enhance access to timely medical advice based on the data that RPM devices produce. In this paper, we report on three studies investigating how the severity of users' medical condition (normal vs. high blood pressure), security risk (low vs. modest vs. high risk), and medical advice source (human doctor vs. AI) influence user perceptions of advisor trustworthiness and willingness to disclose RPM-acquired information. We found that trust mediated the relationship between the advice source and users' willingness to disclose health information: users trust doctors more than AI and are more willing to disclose their RPM-acquired health information to a more trusted advice source. However, we unexpectedly discovered that conditional on trust, users disclose RPM-acquired information more readily to AI than to doctors. We observed that the advice source did not influence perceptions of security and privacy risks. We conclude by discussing how our findings can support the design of RPM applications.

Race and Privacy in Broadcast Police Communications

Radios are essential for the operations of modern police departments, and they function as both a collaborative communication technology and a sociotechnical system. However, little prior research has examined their usage or their connections to individual privacy and the role of race in policing, two growing topics of concern in the US. As a case study, we examine the Chicago Police Department's (CPD's) use of broadcast police communications (BPC) to coordinate the activity of law enforcement officers (LEOs) in the city. From a recently assembled archive of 80,775 hours of BPC associated with CPD operations, we analyze human-generated text transcripts of radio transmissions broadcast 9:00 AM to 5:00 PM on August 10th, 2018 in one majority Black, one majority White, and one majority Hispanic area of the city (24 hours of audio) to explore four research questions: (1) Do BPC reflect reported racial disparities in policing? (2) How and when is gender, race/ethnicity, and age mentioned in BPC? (3) To what extent do BPC include sensitive information, and who is put at most risk by this practice? (4) To what extent can large language models (LLMs) heighten this risk? We explore the vocabulary and speech acts used by police in BPC, comparing mentions of personal characteristics to local demographics, the personal information shared over BPC, and the privacy concerns that it poses. Analysis indicates (a) policing professionals in the city of Chicago exhibit disproportionate attention to Black members of the public regardless of context, (b) sociodemographic characteristics like gender, race/ethnicity, and age are primarily mentioned in BPC about event information, and (c) disproportionate attention introduces disproportionate privacy risks for Black members of the public. This study shows BPC can provide a novel window into disproportionate attention (i.e., via radio communications) by law enforcement officers to specific racial groups, leading to increased privacy vulnerability for those groups, particularly Black males.

The Impact of Artificial Intelligence on Economic Development: A Systematic Review

Artificial Intelligence (AI) has emerged as a transformative force across various sectors, reshaping economies and societies globally. This review aims to provide a systematic analysis of the existing literature on the impact of AI on economic development. By conducting a combined bibliometric and content analysis of relevant studies from the past two decades, we identify major themes and research directions within the field. The review reveals that AI plays a significant role in enhancing productivity, fostering innovation, and driving economic growth. Key areas of influence include intelligent decision-making, labor market transformations, Industry 4.0, and social governance. However, AI also presents challenges such as ethical concerns, potential job displacement, and privacy risks. The findings highlight both opportunities and limitations of AI in economic contexts, emphasizing the need for policies that support positive economic impacts while mitigating adverse effects. This review provides scholars, policymakers, and industry leaders with a comprehensive understanding of AI’s evolving role in economic development and outlines future research directions to bridge current knowledge gaps.

Societal Attitudes Toward Service Robots: Adore, Abhor, Ignore, or Unsure?

Societal or population-level attitudes are aggregated patterns of different individual attitudes, representing collective general predispositions. As service robots become ubiquitous, understanding attitudes towards them at the population (vs. individual) level enables firms to expand robot services to a broad (vs. niche) market. Targeting population-level attitudes would benefit service firms because: (1) they are more persistent, thus, stronger predictors of behavioral patterns and (2) this approach is less reliant on personal data, whereas individualized services are vulnerable to AI-related privacy risks. As for service theory, ignoring broad unobserved differences in attitudes produces biased conclusions, and our systematic review of previous research highlights a poor understanding of potential heterogeneity in attitudes toward service robots. We present five diverse studies (S1–S5), utilizing multinational and “real world” data (Ntotal = 89,541; years: 2012–2024). Results reveal a stable structure comprising four distinct attitude profiles (S1–S5): positive (“adore”), negative (“abhor”), indifferent (“ignore”), and ambivalent (“unsure”). The psychological need for interacting with service staff, and for autonomy and relatedness in technology use, function as attitude profile antecedents (S2). Importantly, the attitude profiles predict differences in post-interaction discomfort and anxiety (S3), satisfaction ratings and service evaluations (S4), and perceived sociability and uncanniness based on a robot’s humanlikeness (S5).

Investigating the Drivers of E-Commerce Continuance Intentions Among Older Consumers in Latin America

This study investigates the drivers of e-commerce continuance intentions among older consumers in Latin America. It employs the Technology Acceptance Model (TAM) and the Unified Theory of Acceptance and Use of Technology (UTAUT2) to develop and test a conceptual model of e-commerce adoption for older adults in Chile. Data were collected through an online survey involving 202 participants aged 65 and over. The findings reveal that attitude toward e-commerce and three UTAUT2 predictors—performance expectancy, effort expectancy, and social influence—are significantly related to e-commerce continuance intentions. Consumer innovativeness also has an indirect relationship through its impact on attitudes toward e-commerce. Additionally, facilitating conditions and habit are indirectly and significantly related to e-commerce continuance intentions through their influence on effort expectancy. However, hedonic motivation, perceived privacy risk, and health condition do not serve as significant predictors of e-commerce continuance intentions. This study enhances the understanding of older consumers’ intentions to continue adopting e-commerce in a Latin American country.

Privacy-Preserving Techniques in Generative AI and Large Language Models: A Narrative Review

Generative AI, including large language models (LLMs), has transformed the paradigm of data generation and creative content, but this progress raises critical privacy concerns, especially when models are trained on sensitive data. This review provides a comprehensive overview of privacy-preserving techniques aimed at safeguarding data privacy in generative AI, such as differential privacy (DP), federated learning (FL), homomorphic encryption (HE), and secure multi-party computation (SMPC). These techniques mitigate risks like model inversion, data leakage, and membership inference attacks, which are particularly relevant to LLMs. Additionally, the review explores emerging solutions, including privacy-enhancing technologies and post-quantum cryptography, as future directions for enhancing privacy in generative AI systems. Recognizing that achieving absolute privacy is mathematically impossible, the review emphasizes the necessity of aligning technical safeguards with legal and regulatory frameworks to ensure compliance with data protection laws. By discussing the ethical and legal implications of privacy risks in generative AI, the review underscores the need for a balanced approach that considers performance, scalability, and privacy preservation. The findings highlight the need for ongoing research and innovation to develop privacy-preserving techniques that keep pace with the scaling of generative AI, especially in large language models, while adhering to regulatory and ethical standards.

Privacy and personal data risk governance for generative artificial intelligence: A Chinese perspective

The rapid development of generative artificial intelligence (AI) has attracted global attention and posed challenges to existing data governance frameworks. The increased technical complexity and expanded scale of data usage not only make it more difficult to regulate AI but also present challenges for the current legal system. This article, which takes ChatGPT's training data and working principles as a starting point, examines specific privacy risks, data leakage risks, and personal data risks posed by generative AI. It also analyzes the latest practices in privacy and personal data protection in China. This article finds that while China's governance on privacy and personal data protection takes a macro-micro integration approach and a private-and-public law integration approach, there are shortcomings in the legal system. Given that the current personal data protection system centered on individual control is unsuitable for the modes of data processing by generative AI, and that private law is insufficient in safeguarding data privacy, urgent institutional innovation is needed to achieve the objective of “trustworthy AI.”

A starting framework for urban AI applications

There has been a growing concern about the ethical implications of Artificial Intelligence (AI) in urban environments. Scholars have highlighted the dual nature of AI, which can both enhance urban sustainability and pose significant privacy risks. The literature acknowledges that AI introduces specific challenges in urban contexts. However, there remains a gap in comprehensive sector-specific ethical frameworks tailored to the unique challenges posed by AI applications in cities. This study aims to fill this gap by proposing a starting framework for the ethics of Urban AI applications, providing principles to guide responsible AI development and deployment in urban settings. The study employed a two-phase methodology: a content analysis of 100 AI ethics documents to develop a preliminary framework, followed by a survey of experts and stakeholders to validate and expand this framework by identifying urban-specific risks and ethical considerations. The developed ethical framework for Urban AI applications provides a starting point to pave the way for ethical and sustainable urban AI systems that enhance the quality of life for all city residents.

Hybrid Prompt Learning for Generating Justifications of Security Risks in Automation Rules

Trigger-action platforms (TAPs) enable users without programming experience to personalize the behavior of Internet of Things applications and services through IF-THEN rules. Unfortunately, the arbitrary connection of smart devices and online services, even with simple rules, such as “IF the entrance Netatmo Wheather Station detects a temperature above 30 \({}^{\circ}C\) ( \(86^{\circ}F\) ) THEN open the shutters in the living room,” might expose users to potential security and privacy risks (e.g., the execution of the previous rule might provide an easy entry point for thieves, especially during the summer vacation period). The goal of our research is to make the users capable of understanding and mitigating the threats and risks associated with the execution of IF-THEN rules. To this end, we define a new challenging task, namely generating post hoc justifications of privacy and security risks associated with automation rules, and propose a novel natural language generation strategy based on hybrid prompt learning producing justifications in the form of real-life threat scenarios. The proposed strategy allows for prompt customization with task-specific information, providing contextual details enabling to grasp the nuances and subtleties of the domain language, resulting in more coherent justifications. The experiments conducted on the if-this-then-that (IFTTT) platform show that our method produces effective justifications, improving the explainability of discrete and hybrid prompting methods up to 27% in BLEURT score. The code of the software is publicly available on GitHub 1 .