Research
I test assumptions and let evidence guide the work, from dark mode displays to anomaly alerts.
Publication
Dark mode vogue: Do light-on-dark displays have measurable benefits to users?
Dark mode is a popular design trend, yet most prior research favors light mode for human performance. I studied whether dark mode’s popularity comes from its aesthetics or from real physiological benefits, comparing younger and older adults on writing and search tasks in bright and dim rooms. Dark mode increased cognitive load for older adults in bright rooms and for younger adults in dim rooms, and older adults also reported more mental effort with it when searching in a bright room. Older adults felt more positive using light mode, while younger adults showed more interest in dark mode, along with more boredom. The findings suggest older adults use light mode to avoid mental fatigue, while younger adults prefer dark mode in dim settings for aesthetic reasons.
- Design
- Mixed design with four groups: age group and lighting between subjects, display and task within subjects
- Participants
- 40 adults: 20 younger (ages 18 to 33) and 20 older (ages 45 to 60), each age group split evenly between lighting conditions
- Conditions
- Display: dark mode vs. light mode
Lighting: bright room (350 to 355 lux) vs. dim room (10 to 15 lux) - Tasks
- Writing and search
- Measures
- Eye tracking: search time, pupil diameter, and fixation count
Self-reported: mental effort, task difficulty, and emotion - Published
- Ergonomics, 2023, from my master’s thesis
Patent
Anomaly Detection at Scale
The idea came from research I conducted with internal subject matter experts on the performance anomaly monitoring experience I designed for Dell CloudIQ (now Dell AIOps). They described a real need to reduce alert fatigue and to make anomaly detection more accurate, easier to configure, and able to learn from user feedback.
This disclosure describes how machine learning can improve performance anomaly detection accuracy and reduce alert fatigue associated with false-positive anomaly determinations. The algorithm 1) compares anomalous behavior with external factors (such as natural disasters, supplier outages, and civil disruptions) to diagnose performance fluctuations and provide recommended actions, and 2) collects user input about the enterprise and allows the correction of false-positive anomalies.
- Problem
- Alert fatigue from false-positive anomaly alerts
- Method
- Machine learning using external context and user feedback
- My Role
- Conceived the original idea, based on anomaly detection research
- Organization
- Dell Technologies
- Status
- U.S. patent pending, 2023
Earlier Writing
Analyses of real products from my master’s program.
- Pre-attentive Processing: how the visual system sorts information in under half a second, applied to a review of the Fidelity app. Read the article (PDF)
- Supporting Metacognitive Decision-making: how designers can reduce choice overload, applied to 1-800 Flowers. Read the article (PDF)
- Memory Load and Emotion: how working memory limits affect task performance, applied to a state vaccine-scheduling site. Read the article (PDF)