The COVID-19 pandemic necessitated companies’ unprecedented digital transformation of their conventional working arrangements by enabling remote e-working (REW) practices supported by information and communication technologies (ICTs). Consequently, working remotely using multiple technologies is the postpandemic new norm for many employees all over the globe. This shift toward remote work can potentially cause technostress (TS) due to the constant use of ICT. With an abundance of research related to the challenges and impacts of REW, less is known about how employees might use coping skills as a useful tool to lessen the negative effects of TS and decrease employee remote work exhaustion (RWE). Based on the job demand–resource (JD-R) theory, the study’s objective is to explore the relationship between REW and remote work. The data of 291 respondents was collected from engineering professionals working in the IT industries of Pakistan, selected based on purposive sampling. The study demonstrates the impact of TS as a mediator and problem-focused coping strategies (CSs) as a moderator, emphasizing their role in the employee level of exhaustion while working remotely. The results of this research add to the current knowledge by providing insights into the significance of positive REW experience by reducing RWE within Pakistan’s IT sector. From an academic standpoint, this study enhances REW research by exploring the consequences of lived experiences and confirming TS as a major stressor that increases exhaustion, and finds that problem-focused coping, though intended as a resource, ultimately exacerbates exhaustion. On a managerial level, the findings provide valuable insights to organizations operating in these sectors, enabling the employees to formulate CSs that are effective at the organizational level and establish formal REW practices and policies to mitigate TS and RWE, leading to an improvement in employees’ well-being.
Dark patterns are not just bad design. They are manipulative interface mechanisms that can steer people’s choices and create real harm, especially in regulated industries, where they can trigger legal, financial, and reputational risks. To help organizations review risk before release at the level of individual screens, we analyze 164 real-world dark-pattern reports collected through the Dark Patterns Tip Line, a public reporting initiative hosted by Stanford PACS’ Digital Civil Society Lab. We propose a Responsible AI workflow that connects a five-category taxonomy to AI-based detection and then turns model outputs into practical decision aids for governance. We compare two deployment-realistic analysis routes: 1) a vision + law + LLM route that reads the screen as an image and can capture visual cues (e.g., layout, defaults, and salience), and 2) an OCR/text-forward route that relies mainly on extracted text. Instead of focusing only on accuracy, we emphasize reliability: the model’s confidence should match real-world error rates. We, therefore, evaluate probability calibration using expected calibration error (ECE) and multiclass Brier score while keeping overall classification performance comparable. Results indicate two key issues: both routes show a systematic tendency to over-label Interface Interference, which can distort governance priorities, and in this corpus, the vision-forward route provides substantially better-calibrated probabilities (ECE ≈ 0.19 versus ≈ 0.52; Brier ≈ 0.258 versus ≈ 0.489). Better calibration makes model-generated category assignments more reliable before they are combined with the empirical likelihood (L) and policy-grounded impact (I) scores, thereby reducing unnecessary “very high” escalations.
