Deskly: A Privacy-First Desktop Digital Wellbeing System for Windows Using Behavioral Nudges and Gamification

Key findings from this study

• The study found that Deskly reduced daily screen time by 23% among 12 participants with a large effect size (Cohen's d = 0.93).
• The authors report a 17% improvement in WHO-5 Well-Being Index scores following system deployment (p = 0.036, d = 0.69).
• The researchers demonstrate that 100% of participants adopted the graduated extension model requiring conscious acknowledgment before extending app time limits.
• The study found that application switching frequency decreased by 22% per hour during the evaluation period (d = 0.76).

Overview

Deskly is a desktop-native digital wellbeing system developed for Microsoft Windows to address the absence of intervention-capable wellbeing tools on desktop platforms, which serve as primary work devices for students and knowledge workers. The system integrates real-time application and browser-level activity tracking through Windows UI Automation APIs, behavioral nudge interventions with graduated app limits and conscious-extension checkpoints, gamified habit formation mechanisms, and privacy-first local data storage using an Obsidian-inspired vault architecture. The technical implementation uses Electron 31, React 18, and Vite 5, with PowerShell-C# interop enabling sub-second foreground detection including UWP apps. Browser title-bar parsing achieves website-level tracking across 58 browsers without requiring extensions. Heuristic app categorization employs 88 keyword rules and 7 intent labels, achieving a weighted F1 score of 0.83. Local-only JSON vault storage incorporates auto-repair and 60-second write-behind caching. The system was developed at Parul University, Vadodara, Gujarat, India.

Methods and approach

The system architecture employs Windows UI Automation APIs for real-time activity monitoring at both application and browser levels. PowerShell-C# interoperability enables sub-second foreground window detection including Universal Windows Platform applications. Website-level tracking across 58 browsers occurs through title-bar parsing without browser extension dependencies. Application categorization uses a heuristic model with 88 keyword rules distributed across 7 intent labels. Privacy preservation relies on local-only JSON vault storage with auto-repair functionality and 60-second write-behind caching. Behavioral interventions implement graduated app limits with conscious-extension checkpoints requiring user acknowledgment. Gamification mechanisms include streak tracking, achievement systems, and daily goal structures. The system underwent pilot evaluation using a within-subjects design with 12 participants. Outcome measures included daily screen time, WHO-5 Well-Being Index scores, System Usability Scale ratings, adoption rates of graduated extension features, and application switching frequency.

Results

The within-subjects pilot evaluation with 12 participants demonstrated a 23% reduction in daily screen time (p = 0.008, Cohen's d = 0.93). Participants showed a 17% improvement in WHO-5 Well-Being Index scores (p = 0.036, d = 0.69). The system achieved a System Usability Scale rating of 81.3 out of 100, classified as excellent. All participants adopted the graduated extension model, indicating 100% uptake of the conscious-extension checkpoint feature.

Application switching behavior decreased by 22% per hour (d = 0.76). The heuristic app categorization system achieved a weighted F1 score of 0.83 using 88 keyword rules across 7 intent categories. Browser title-bar parsing successfully tracked website-level activity across 58 different browsers without requiring browser extensions. PowerShell-C# interoperability enabled foreground detection with sub-second latency for both traditional and Universal Windows Platform applications. The local JSON vault architecture with 60-second write-behind caching maintained data integrity while preserving privacy through local-only storage.

Implications

The findings establish feasibility for desktop-native digital wellbeing interventions on Windows platforms, addressing a gap in existing tools predominantly designed for mobile devices. The 23% reduction in screen time with large effect size suggests that behavioral nudges combining graduated limits and conscious-extension checkpoints effectively modify desktop usage patterns among students and knowledge workers. The 17% improvement in wellbeing scores indicates that desktop-focused interventions can influence broader psychological outcomes beyond simple usage reduction. Complete adoption of the graduated extension model demonstrates user acceptance of friction-based intervention designs when implemented with appropriate scaffolding.

The technical approach demonstrates that privacy-preserving architectures using local-only storage can achieve functionality comparable to cloud-based alternatives while maintaining user data sovereignty. Browser title-bar parsing eliminates extension dependency, reducing deployment friction and compatibility issues across diverse browser ecosystems. The heuristic categorization approach with 0.83 weighted F1 score suggests rule-based systems can achieve acceptable accuracy for intent classification without machine learning infrastructure requirements. The 22% reduction in application switching suggests that usage awareness and intervention mechanisms may reduce context-switching behavior associated with productivity losses and cognitive load.

Scope and limitations

This summary is based on the study abstract and available metadata. It does not include a full analysis of the complete paper, supplementary materials, or underlying datasets unless explicitly stated. Findings should be interpreted in the context of the original publication.