CASE 06 / UX RESEARCH / 2026
The Kind Design Paradox
The paradox of
helpful design.
Does more structure lead to better understanding?
A helpful-looking hierarchy does not guarantee understanding. Brief notifications need information design that considers how people actually read them.
- PARTICIPANTS
- 125
- STUDY DESIGN
- 2 × 2
- ROLE
- First author
SAME INFORMATION, DIFFERENT DESIGN
Same information, different structure. Original stimuli A and D.
- QUESTION
- Does organizing information also improve understanding?
- APPROACH
- Four message structures, 125 participants, comprehension and cognitive load.
- TAKEAWAY
- Design for the reading context and next action before adding more hierarchy.
01 / RESEARCH QUESTION
The information was there. The questions kept coming.
Remote-class notices contained the joining instructions, materials, and schedule. Yet the same questions kept returning. Providing information and helping people understand and act on it were different problems.
“Would clearer grouping and emphasis solve it?”
Chunking and visual priority are familiar design principles. I tested whether their expected benefits would carry over to brief notification messages.
02 / EXPERIMENT
Same information. Four structures.
Using a fictional English-class notice, I varied chunking and visual priority in a 2 × 2 between-subjects experiment. Message bodies contained 187–190 characters.
Plain text
Chunking Low / Priority Low · n=31
Visual emphasis
Chunking Low / Priority High · n=30
Chunking
Chunking High / Priority Low · n=31
Chunking + emphasis
Chunking High / Priority High · n=33
Original study screens from Figure 2. The blue “Next” button advances the experiment; it is separate from the message CTA. Visual priority combined emoji and CTA presence/placement.
03 / MEASUREMENT
After reading, what did people understand?
- 01
Read a message
View one randomly assigned condition
- 02
Test understanding
Five comprehension items and confidence
- 03
Measure effort
Mean of six raw NASA-TLX subscales
Sample and method details
Of 132 completed responses, four incomplete records, one response-time outlier, and two repeated-choice responses were excluded. The final sample was 125; the recommended target was 128. Five pilot participants were excluded from the main study.
Primary outcomes were comprehension accuracy and cognitive load. Analysis used 2 × 2 between-subjects ANOVA, with ANCOVA and nonparametric robustness checks.
04 / FINDINGS
More design did not guarantee better understanding.
Condition D, combining chunking and emphasis, had the numerically lowest accuracy and highest cognitive load. This prompted a closer look at the assumption that clearer structure always improves understanding.
Comprehension accuracy
Higher is better · %
Condition means · full scale 0–100
Cognitive load
Lower is better · NASA-TLX score
Condition means · full scale 0–100
Read the pattern together with the statistical result. Neither primary outcome showed significant main effects or interaction. The differences in D do not establish that chunking or emphasis harms comprehension.
View statistical tests, means and standard deviations
| Effect | F | p | ηp² |
|---|---|---|---|
| Chunking | 0.71 | .402 | .006 |
| Visual priority | 1.27 | .263 | .010 |
| Interaction | 0.47 | .494 | .004 |
| Accuracy: A 85.2 (18.6), B 83.3 (22.3), C 84.5 (25.7), D 77.0 (25.6) · M (SD) | |||
| Effect | F | p | ηp² |
|---|---|---|---|
| Chunking | 2.12 | .148 | .017 |
| Visual priority | 1.40 | .240 | .011 |
| Interaction | 0.06 | .814 | .000 |
| Load: A 40.2 (20.6), B 43.4 (19.4), C 44.3 (19.1), D 49.0 (16.0) · M (SD) | |||
Secondary outcomes and interpretation limits
Confidence (1–7) was A 5.06, B 5.03, C 4.68, D 4.15. Calibration gaps were A 20.9, B 21.2, C 24.5, D 34.3; D had the numerically largest gap between self-assessment and performance.
Some calibration analyses had nominal p<.05, but none met the Bonferroni threshold for six comparisons (.0083). These remain exploratory signals. A ceiling effect (52% scored full marks), one education message, and combined visual manipulations limit interpretation.
Message length, exposure duration, and one-off reading context are conditions to isolate in future studies.
05 / DESIGN IMPLICATIONS
Design for a brief encounter, not just a tidy message.
The study moves the design question from how much to emphasize toward what to deliver, where, and when. Information volume and delivery stages need to be designed together.
- WHAT
Make the immediate action clear
Focus the notification on the action and essentials needed now.
- WHERE
Give details room to be read
Explore moving rules, preparation, and details to a landing page.
- WHEN
Deliver in relevant stages
Explore staged guidance at signup, preparation, and before class.
These are follow-up design hypotheses proposed by the paper. Their effects require separate validation.
06 / IN PRACTICE
In practice, connect information to action.
The same design question appears in a real class notice: moving from several pieces of information in one message toward a clearer connection between the current situation and the next action.
Several instructions at once
Readers had to locate the relevant information and action.
From the current situation to the next step
The notice connects the class information more directly to the entry action.
Operational observation and its limits
An operator count recorded recurring inquiries moving from 12 per month to 0–1 in the first half of 2026. Message volume and the comparison-month basis are not specified. This is an operational observation, not an experimental or causal result of the paper.
07 / CONTRIBUTION & PUBLICATION
From applying a principle to asking when it applies.
As first author, I led the study design and analysis. I translated an operational question into testable variables, then used unexpected patterns to frame further questions about context.
My design criterion is clear: look beyond an orderly presentation and check whether people can actually understand and act.