Introduction
In the field of Human-Computer Interaction (HCI), digital tools are predominantly designed for functional efficiency—optimizing tasks like navigation, data retrieval, or logistics. However, many human activities are "emotional" searches where the process is as significant as the outcome. Finding a four-leaf clover (FLC) is a quintessential example: a global symbol of luck with a natural occurrence rate of approximately 1 in 10,000.
This research introduces "Happiness Finder," an AI-powered system designed to assist users in this search. The study explores a fundamental tension: does AI assistance diminish the "spark" of discovery and human agency, or can it enhance well-being? The authors propose the concept of "Slow Digital"—technology designed to promote mindful engagement with the physical world rather than mere productivity.
System Architecture and Methodology
The researchers developed a web-based application that utilizes a mobile-to-cloud pipeline to provide real-time feedback. To avoid the "novelty effect" associated with high-end AR headsets, the system was deployed on standard smartphones and tablets.
As shown in the architecture above, the process involves four key steps:
- Capture: The mobile device captures video frames of clover patches.
- Inference: Frames are transmitted at one-second intervals to an AWS cloud server.
- Detection: The server employs a YOLOv8x object detection model to identify FLCs.
- Feedback: Upon high-confidence detection, the server returns coordinates for a Bounding Box (BBox) and triggers a notification sound on the user's device.
Dataset and Model Performance
The YOLOv8x model was trained using a specialized FLC dataset sourced from a benchmark for fine-grained object localization (Bravo et al., 2019). The training set consisted of 1,234 images, with 310 images reserved for evaluation. Using the Roboflow annotation tool, the model achieved a precision of 0.676, a recall of 0.500, and a mean Average Precision ([email protected]) of 0.533.
User Interaction and the "Slow Digital" Experience
The core of the study focused on how AI intervention affects the user's emotional state. By superimposing digital markers over the physical world, the system acts as a guide rather than a replacement for human observation.
This interaction model supports the "Slow Digital" philosophy, encouraging users to spend time outdoors and engage with nature. The researchers found that even when the AI performed the detection, users maintained a high level of psychological reward.
Key Results and Insights
A pilot study involving 22 participants from diverse geographic backgrounds yielded several significant findings:
- Positive Sentiment: 68.1% of participants preferred the AI-assisted search over traditional manual searching.
- Emotional Impact: An overwhelming 90.9% of users reported feeling happy upon finding an FLC with the system.
- Attribution of Success: Interestingly, 66.7% of users credited the system for the find, while only 12.5% felt they found it independently. Despite this "outsourcing" of the discovery, the joy associated with the find remained intact.
- Cultural Context: The study noted that the emotional "Return on Investment" (ROI) is culturally dependent; while FLCs are potent symbols of luck in Western cultures, awareness was lower among participants from parts of Asia and the Middle East.
Discussion and Future Work
The findings suggest that AI can facilitate "hidden treasure" searches without stripping away the emotional value of the discovery. However, the study was limited by its small sample size and the use of controlled environments (potted artificial clovers).
Future research will explore varying levels of AI support. Instead of providing a direct bounding box, the system could offer "hot or cold" hints to preserve more human agency and the "thrill of the hunt." Additionally, the researchers aim to test the system in wild, natural environments to assess long-term engagement and the sustainability of the happiness boost provided by AI assistance.
Summary updated