5D Library › Computational Aesthetics, Visual Saliency Maps and Cognitive Ergonomics in User Interface Design · 中文

Chapter 8 Practice and a Pre-Registered Test

Note

The checklist and trial protocol in this chapter are for research and self-checking; they are not a design standard or business or compliance advice. The checklist cannot tell you how much conversion a page will bring; it can only help you write down your assumptions before you make a change. All pages, brands and data in the book are fictional or simulated.

Part 1 A four-step checklist

Chapters 1–6 can be condensed into a four-step checklist for examining an interface before and after a change. The limits in it (for example “the target should be among the model’s first three shifts of attention”, “a budget of 1,500 ms for deciding and pointing”) are examples of rules a user should write down in advance, not recommended values.

  1. Salience. Computing the page with a saliency model (Chapter 3), is the task target among the first few shifts of attention? Remember that the model does not know the task, does not read text and has no centre bias (Chapter 4); this step can only catch obvious problems, such as a target that differs from its surroundings in none of colour, intensity and orientation.
  2. Legibility. Does every pair of text and background reach the WCAG 2.x contrast requirements (4.5:1 for body text, 3:1 for large text)? Is the x-height of small text below about 0.2 degrees at the expected viewing distance (Chapter 5)?
  3. Choice and pointing. Computing the uncertainty of the decision from the historical or expected click distribution (Hyman’s entropy, Chapter 2), plus the Shannon-form Fitts difficulty to the target, is the total time within a budget written down in advance? For novices, the cost of more options is roughly linear, not logarithmic.
  4. Evidence. Does any claim that “conversion rose after the change” come from a pre-registered test with a single primary outcome and a correction for multiple comparisons (Chapter 1)? If not, treat it as untested.

The program below applies the checklist to a fictional landing page (layout B of Chapter 3, with a grey call-to-action button and light-grey card labels) and its revision (layout A, with a red button and darker card labels).

Calculation

# The four-step checklist of Chapter 8 applied to the fictional landing page of p06
# (layout B, grey CTA) and a revised version (layout A, red CTA). Limits are examples
# of rules written down BEFORE looking, not recommendations.
#  1 salience: is the task target among the first 3 attended locations of the model?
#  2 legibility: does every text/background pair reach WCAG AA (4.5:1; 3:1 if large)?
#  3 choice and pointing: expert decision (Hick-Hyman over the click distribution) +
#    Shannon-Fitts time to the CTA (300 px away, 60 px wide), budget 1,500 ms
#  4 evidence: is a claimed improvement backed by a pre-registered test?
import math, importlib.util, io, contextlib, pathlib, sys
sys.dont_write_bytecode = True

def load(name):
    spec = importlib.util.spec_from_file_location(name, pathlib.Path(__file__).with_name(
        name + ".py"))
    m = importlib.util.module_from_spec(spec)
    with contextlib.redirect_stdout(io.StringIO()):
        spec.loader.exec_module(m)
    return m

p06, p10 = load("p06_itti_koch"), load("p10_wcag_contrast")
to8 = lambda c: tuple(round(255 * v) for v in c)
TEXT = {"headline on page": ((.1, .1, .1), (.97, .97, .97), True),
        "body text on page": ((.35, .35, .35), (.97, .97, .97), False),
        "nav links on nav bar": ((.55, .55, .58), (.15, .15, .17), False)}
P_CLICK = [0.55, 0.20, 0.10, 0.06, 0.05, 0.04]          # CTA, 2nd button, 4 cards

def audit(name, cta, label_on_cta, card_label, claim_preregistered):
    print(f"{name}")
    S, _ = p06.saliency(p06.page(cta))
    first = [p06.label(x, y) for x, y, _ in p06.scanpath(S, k=3)]
    ok1 = "CTA button" in first
    print(f"   1 salience   first 3 model shifts: {', '.join(first)}"
          f" -> {'pass' if ok1 else 'FLAG'}")
    pairs = dict(TEXT, **{"CTA label on CTA": (label_on_cta, cta, False),
                          "card labels on cards": (card_label, (.93, .93, .93), False)})
    bad = []
    for k, (fg, bg, large) in pairs.items():
        r = p10.ratio(to8(fg), to8(bg))
        if r < (3.0 if large else 4.5):
            bad.append(f"{k} {r:.2f}:1")
    print(f"   2 legibility {len(pairs) - len(bad)}/{len(pairs)} pairs pass"
          + (f"; FLAG: {'; '.join(bad)}" if bad else " -> pass"))
    H = -sum(p * math.log2(p) for p in P_CLICK)
    t = 200 + 150 * H + 100 + 150 * math.log2(300 / 60 + 1)
    print(f"   3 choice     H = {H:.2f} bits; decide + point to CTA {t:.0f} ms"
          f" -> {'pass' if t <= 1500 else 'FLAG'}")
    print(f"   4 evidence   claimed lift pre-registered: "
          f"{'yes -> pass' if claim_preregistered else 'no -> FLAG (treat as untested)'}")

if __name__ == "__main__":
    audit("fictional page, layout B (grey CTA, light-grey card labels)", (.35, .35, .35),
          (.97, .97, .97), (.6, .6, .6), False)
    audit("revised page, layout A (red CTA, darker card labels)", (.85, .1, .1),
          (1.0, 1.0, 1.0), (.4, .4, .4), False)
    print("note: step 1 is a model prediction, not an eye-tracking result; step 4 stays")
    print("   flagged until the Chapter 8 trial (or an equivalent test) has been run")
# fictional page, layout B (grey CTA, light-grey card labels)
#    1 salience   first 3 model shifts: logo, hero photo, hero photo -> FLAG
#    2 legibility 4/5 pairs pass; FLAG: card labels on cards 2.43:1
#    3 choice     H = 1.92 bits; decide + point to CTA 975 ms -> pass
#    4 evidence   claimed lift pre-registered: no -> FLAG (treat as untested)
# revised page, layout A (red CTA, darker card labels)
#    1 salience   first 3 model shifts: CTA button, logo, CTA button -> pass
#    2 legibility 5/5 pairs pass -> pass
#    3 choice     H = 1.92 bits; decide + point to CTA 975 ms -> pass
#    4 evidence   claimed lift pre-registered: no -> FLAG (treat as untested)
# note: step 1 is a model prediction, not an eye-tracking result; step 4 stays
#    flagged until the Chapter 8 trial (or an equivalent test) has been run

The original page is flagged at step 1: the model’s first three shifts of attention are the logo, the photograph and the photograph again, and the call-to-action button is not among them. It is also flagged at step 2: of five text pairs, the card labels on the card background reach only 2.43:1, below 4.5:1. Step 3 passes: with the assumed click distribution, the uncertainty of the decision is 1.92 bits, and deciding plus pointing to the button takes about 975 ms, within the budget of 1,500 ms. Step 4 is flagged: there is no pre-registered test. The revision passes steps 1 and 2 (the button becomes the first shift of attention and all five text pairs reach 4.5:1), and step 3 is unchanged; step 4 is still flagged—passing the checklist does not mean the change works. Whether the revision really lets users complete tasks faster can be answered only by the trial below.

Part 2 A pre-registered online randomised controlled trial

Question and design choices

On a set of e-commerce pages for fictional brands, do layouts revised by this book’s checklist (raising the salience of the target and the contrast of text, and shortening the Fitts distance to the target) shorten the time to “find and correctly click a specified target” compared with the original layouts, by at least the smallest effect of interest?

The author’s framework proposes comparing “layouts optimised for saliency, contrast and Fitts distance” with “control layouts”. The book keeps the comparison and makes four design choices. First, a parallel-group design: each participant sees only one version of the pages, because someone who has seen one layout will be influenced by memory when looking for the same target in the other. Second, a single primary outcome: the “log time to a correct click”, rather than “first-click accuracy”. The reasons: completion time is continuous and gives more power for the same sample size; it corresponds directly to the quantity predicted by the models of Chapters 2–4 (the sum of search, decision and pointing times); and the log transform makes right-skewed reaction times roughly normal, so the difference between groups can be read as a proportional change in geometric-mean time. First-click accuracy is a secondary outcome and also a pre-specified qualifier: if the optimised layouts are faster but clearly less accurate, the conclusion must say so. Third, covariate adjustment: before seeing the experimental layouts each participant does two neutral practice tasks, whose log time is the covariate in an analysis of covariance, absorbing individual differences in speed (Borm et al. 2007). Fourth, implementation by independent researchers: page construction, randomisation, data collection and analysis are all done by researchers unconnected with the author’s business; the author supplies only the checklist and revision rules published in this book.

Protocol

Element Specification
Implementers and recruitment Run by a university research team or research institute independent of the author’s business; adults recruited through an online participant pool unconnected with the author; no recruitment from the author’s community, students, customers or the users of affiliated companies, screened out by questionnaire at sign-up
Materials E-commerce pages for 12 fictional brands (no real brand, trademark or page of the author’s companies is used); one specified target per page (such as “add this item to the basket” or “open the returns policy”); two versions: X = revised by the Part 1 checklist (salience, contrast, Fitts distance), Y = original layout; text and target labels identical in both versions; all pages, tasks and revision rules published at registration
Device Desktop browsers only (window at least 1,280 pixels wide); conclusions about mobile devices need a separate trial
Randomisation and stratification The participant platform allocates 1:1 to X or Y, stratified by age band (under 40 / 40 and over), in randomly permuted blocks of 4 or 6; the allocation sequence is generated and sealed by someone not involved in the analysis
Blinding Participants are told this is a “website usability study”, without being told that there are two versions or what the hypothesis is; until results are locked, analysts see groups only as coded “A / B”
Primary outcome Each participant’s mean over 12 tasks of ln\ln (seconds from page display to a correct click on the target); tasks not completed within 60 s are counted as 60 s; X − Y compared by analysis of covariance with the mean log time of the two practice tasks as covariate, reporting the difference with its 95 % confidence interval
Smallest effect of interest A 10 % shorter geometric-mean completion time, δ=ln⁡0.90=−0.105\delta = \ln 0.90 = -0.105
Pre-specified qualifier Whether X’s first-click accuracy is more than 3 percentage points below Y’s; if so, the conclusion says so
Secondary outcomes First-click accuracy; completion within 60 s; number of misclicks; System Usability Scale (SUS) score (Brooke 1996); all descriptive or exploratory, never changing the primary conclusion

Guarding against data dredging

There is a single primary outcome and a single pre-specified analysis model; pages, revision rules, randomisation scheme, exclusion criteria and analysis code are registered publicly before recruitment, with the hash of the code registered alongside (Nosek et al. 2018); there is no interim analysis and no data-driven change to exclusion criteria; secondary and exploratory results are clearly labelled and cannot substitute for the primary conclusion. Book 33 discussed the statistical rationale for these practices.

Sample size

Sample size requires assumptions about effect and variance. No prior study can be borrowed directly, so the values below are all planning assumptions, which the editorial notes ask the author to confirm: a between-person standard deviation of log completion time of 0.30, a within-person (between-task) standard deviation of 0.50, a geometric-mean time per task of 8 s in the control group, and a correlation of 0.5 between practice tasks and the primary outcome. The standard deviation of the mean over 12 tasks is then 0.302+0.502/12=0.333\sqrt{0.30^2 + 0.50^2/12} = 0.333. In the normal approximation, the number of analysable participants needed per group is

n=2(z0.975+z0.90)2σ2(1−ρ2)δ2,n = \frac{2\,(z_{0.975} + z_{0.90})^2\,\sigma^2\,(1-\rho^2)}{\delta^2},

where (1−ρ2)(1-\rho^2) is the variance reduction of analysis of covariance relative to a simple tt-test (Borm et al. 2007).

Calculation

# Sample size for the Chapter 8 pre-registered online trial. Two parallel arms, 1:1:
# X = layouts revised by the checklist (salience, contrast, Fitts distance), Y = control
# layouts; 12 find-and-click tasks on fictional shop pages. Single primary outcome: each
# participant's mean of ln(time to the correct click, s), a task not completed within
# 60 s counted as 60 s; analysed by ANCOVA on the mean ln time of 2 neutral practice
# tasks done before the layouts differ. SESOI: 10 % shorter geometric-mean time,
# delta = ln 0.90. PLANNING ASSUMPTIONS (not estimates from data): between-person sd of
# ln time 0.30, within-person task sd 0.50, practice-outcome correlation 0.5.
import math, random, statistics

Z = statistics.NormalDist().inv_cdf
DELTA, SB, SW, RHO, K = math.log(0.90), 0.30, 0.50, 0.5, 12
MU = math.log(8.0)                                 # control geometric mean 8 s per task

def n_arm(delta, sd, rho, power=0.90, alpha=0.05):
    return math.ceil(2 * (Z(1 - alpha / 2) + Z(power)) ** 2 * sd ** 2 * (1 - rho ** 2)
                     / delta ** 2)

def participant(rng, arm):
    b = rng.gauss(0, SB)                           # person effect
    pr = statistics.fmean(MU + b + rng.gauss(0, SW) for _ in range(2))
    shift = DELTA if arm else 0.0
    y = statistics.fmean(min(MU + b + shift + rng.gauss(0, SW), math.log(60.0))
                         for _ in range(K))
    return pr, y

def ancova_t(rows):
    x = [r[1] for r in rows]; g = [r[0] for r in rows]; y = [r[2] for r in rows]
    n = len(y); mx, mg, my = statistics.fmean(x), statistics.fmean(g), statistics.fmean(y)
    sxx = sum((a - mx) ** 2 for a in x); sgg = sum((a - mg) ** 2 for a in g)
    sxg = sum((a - mx) * (b - mg) for a, b in zip(x, g))
    sxy = sum((a - mx) * (c - my) for a, c in zip(x, y))
    sgy = sum((b - mg) * (c - my) for b, c in zip(g, y))
    det = sxx * sgg - sxg ** 2
    bx = (sgg * sxy - sxg * sgy) / det; bg = (sxx * sgy - sxg * sxy) / det
    res = [c - my - bx * (a - mx) - bg * (b - mg) for a, b, c in zip(x, g, y)]
    s2 = sum(r * r for r in res) / (n - 3)
    return bg, math.sqrt(s2 * sxx / det)

if __name__ == "__main__":
    sd = math.sqrt(SB ** 2 + SW ** 2 / K)
    r_pr = SB ** 2 / math.sqrt((SB ** 2 + SW ** 2 / 2) * (SB ** 2 + SW ** 2 / K))
    print(f"assumed sd of the outcome {sd:.3f}; implied practice correlation {r_pr:.2f}"
          f" (planning value {RHO})")
    n0 = n_arm(DELTA, sd, 0.0); n1 = n_arm(DELTA, sd, RHO)
    print(f"normal approximation, 90 % power, two-sided alpha 0.05, delta = ln 0.90"
          f" = {DELTA:.4f}:\n   t-test {n0} per arm; ANCOVA (rho {RHO}) {n1} per arm")
    rng = random.Random(46); hits = 0; sims = 300; est = []
    for _ in range(sims):
        rows = [(arm,) + participant(rng, arm) for arm in (0, 1) for _ in range(n1)]
        b, se = ancova_t(rows); est.append(b)
        hits += b + 1.96 * se < 0                  # CI excludes 0 on the faster side
    print(f"simulation check ({sims} trials of {n1} per arm, lognormal tasks, 60 s cap):"
          f"\n   power {hits / sims:.3f}; mean estimate {statistics.fmean(est):+.4f}")
    rec = math.ceil(n1 / 0.85 / 10) * 10
    print(f"allowing 15 % exclusion/attrition: recruit {rec} per arm, {2 * rec} in total")
    print("sensitivity (per arm, ANCOVA):        SESOI  5 %    10 %    15 %")
    for lab, s_, r_ in (("planning values", sd, RHO), ("sd 30 % higher", 1.3 * sd, RHO),
                        ("no usable covariate", sd, 0.0)):
        row = [n_arm(math.log(1 - e), s_, r_) for e in (0.05, 0.10, 0.15)]
        print(f"   {lab:32s}" + "".join(f"{v:8,d}" for v in row))
# assumed sd of the outcome 0.333; implied practice correlation 0.58 (planning value 0.5)
# normal approximation, 90 % power, two-sided alpha 0.05, delta = ln 0.90 = -0.1054:
#    t-test 210 per arm; ANCOVA (rho 0.5) 158 per arm
# simulation check (300 trials of 158 per arm, lognormal tasks, 60 s cap):
#    power 0.930; mean estimate -0.1055
# allowing 15 % exclusion/attrition: recruit 190 per arm, 380 in total
# sensitivity (per arm, ANCOVA):        SESOI  5 %    10 %    15 %
#    planning values                      664     158      67
#    sd 30 % higher                     1,123     266     112
#    no usable covariate                  886     210      89

A tt-test without the covariate needs 210 per group; with a covariate correlated 0.5, 158 per group. Under the assumed variance structure the actual correlation between practice tasks and the primary outcome is 0.58, above the planning value of 0.5, so planning with 0.5 is conservative; a simulation (300 trials of 158 per group, log-normal task times, a 60 s cap) gives power of 0.930 and a mean effect estimate of −0.1055, matching the true −0.1054. Allowing for 15 % exclusions and dropout, 190 are recruited per group, 380 in all. The sensitivity table shows that a smallest effect of interest of 5 % would need 664 per group; a standard deviation 30 % above the assumption, 266; a practice covariate with no predictive value at all, 210. Figure 5 plots power against the number per group.

Power of the Chapter 8 trial (normal approximation, two-sided \alpha = 0.05) against the number of analysable participants per group. Solid: smallest effect of interest a 10 % shorter geometric-mean time, covariate correlation 0.5; dashed: the same effect without the covariate; dash-dotted: 5 % shorter, covariate correlation 0.5. The vertical line marks the planned 158 per group, the horizontal line 90 % power.

Following Book 33, “90 % power” means that if the true effect exactly equals the smallest effect of interest, the probability of a conclusion such as “X is faster” (upper confidence limit below 0) is about 90 %; in that situation the strongest conclusion, “faster by at least the smallest effect of interest” (upper limit no higher than ln⁡0.90\ln 0.90), has only about a 2.5 % chance of appearing.

Decision rule

Calculation

# Decision rule, fixed before unblinding. Primary outcome: ANCOVA difference X - Y in
# mean ln(time to correct click), adjusted for practice-task ln time; 95 % CI.
# SESOI = ln 0.90 = -0.105 (10 % shorter geometric-mean time). Negative = X faster.
# "No difference worth acting on": CI inside +-0.105 on the log scale, i.e. a time
# ratio between 0.90 and 1/0.90 = 1.111 (-10 % to +11.1 %).
# Gates, checked first: analysable participants below 300 of the 316 planned; arm
# attrition differing by more than 5 points; any page served from the wrong arm on more
# than 1 % of sessions; the analysis code differing from the registered hash
# -> NOT INTERPRETABLE. Pre-specified qualifier (secondary): first-click accuracy lower
# in X by more than 3 percentage points.
import math
SESOI = math.log(0.90)

def verdict(d, lo, hi, n=316, attr_gap=0.01, wrong=0.0, code_ok=True, acc_gap=0.0):
    for bad, why in ((n < 300, f"only {n} analysable (minimum 300)"),
                     (attr_gap > 0.05,
                      f"attrition differs by {100 * attr_gap:.0f} points"),
                     (wrong > 0.01, f"{wrong:.1%} of sessions served the wrong arm"),
                     (not code_ok, "analysis code differs from the registered hash")):
        if bad:
            return f"d = {d:+.3f}: NOT INTERPRETABLE - {why}"
    if hi <= SESOI:       v = "X faster by at least the SESOI (10 %)"
    elif hi < 0:          v = "X faster; not shown to reach the SESOI"
    elif lo > 0:          v = "X slower than control"
    elif lo >= SESOI and hi <= -SESOI: v = "no difference worth acting on (-10 %/+11 %)"
    else:                 v = "unclear: the interval is too wide to decide"
    if acc_gap > 0.03:
        v += f"\n    qualifier: 1st-click accuracy {100 * acc_gap:.0f} points lower in X"
    return f"d = {d:+.3f} ({lo:+.3f} to {hi:+.3f}), time {math.exp(d) - 1:+.0%}: {v}"

for case in ((-0.180, -0.250, -0.110, 316), (-0.090, -0.160, -0.020, 316),
             (-0.010, -0.080, 0.060, 316), (0.080, 0.010, 0.150, 316),
             (-0.060, -0.200, 0.080, 316),
             (-0.200, -0.270, -0.130, 316, .01, 0, True, .05),
             (-0.200, -0.270, -0.130, 281), (-0.200, -0.270, -0.130, 316, 0.08)):
    print(verdict(*case))
print("secondary outcomes (first-click accuracy, completion within 60 s, misclicks,")
print("  SUS score) are reported with intervals; they never decide the verdict")
# d = -0.180 (-0.250 to -0.110), time -16%: X faster by at least the SESOI (10 %)
# d = -0.090 (-0.160 to -0.020), time -9%: X faster; not shown to reach the SESOI
# d = -0.010 (-0.080 to +0.060), time -1%: no difference worth acting on (-10 %/+11 %)
# d = +0.080 (+0.010 to +0.150), time +8%: X slower than control
# d = -0.060 (-0.200 to +0.080), time -6%: unclear: the interval is too wide to decide
# d = -0.200 (-0.270 to -0.130), time -18%: X faster by at least the SESOI (10 %)
#     qualifier: 1st-click accuracy 5 points lower in X
# d = -0.200: NOT INTERPRETABLE - only 281 analysable (minimum 300)
# d = -0.200: NOT INTERPRETABLE - attrition differs by 8 points
# secondary outcomes (first-click accuracy, completion within 60 s, misclicks,
#   SUS score) are reported with intervals; they never decide the verdict

The example results are invented. If any of the preconditions fails (fewer than 300 analysable participants, dropout differing between groups by more than 5 percentage points, more than 1 % of sessions loading pages from the wrong group, or analysis code differing from the registered hash), the report says “not interpretable”, whatever the primary outcome. If the upper confidence limit is no higher than ln⁡0.90\ln 0.90, the report says “X faster by at least the smallest effect of interest”; if the upper limit is below 0 but above ln⁡0.90\ln 0.90, “X faster, size not established”; if the lower limit is above 0, “X slower than control”; if the interval lies within ±ln⁡0.90\pm\ln 0.90 (a time ratio of 0.90 to 1.111, that is −10 % to +11.1 %), “no difference worth acting on”; otherwise “unclear”. The qualifier does not change the category of conclusion; it is appended: in the sixth example, X’s completion time is 18 % shorter but its first-click accuracy is 5 percentage points lower, and the report must state both.

What this trial cannot show

First, it compares a bundle of revisions (salience, contrast and Fitts distance changed together) and cannot attribute the effect to any one of them; separating them requires a factorial design. Second, the primary outcome is the time to find and click a target, not purchase, sign-up or any business outcome; finding a button faster does not mean more people want to buy. Third, the tasks are specified and participants know what they are looking for; the distribution of attention in free browsing may differ (Chapter 4). Fourth, the conclusions apply only to desktop browsers, these 12 fictional pages and the population of the online participant pool; they do not carry over directly to mobile devices, real brands or particular markets. Fifth, it does not test whether the Itti–Koch model’s predictions match fixations (that needs eye-tracking), nor the golden ratio, three clicks or any aesthetic measure. Sixth, the planning assumptions (standard deviations, correlation) are the author’s estimates, not prior data; if the true variance is larger, power will fall below 90 %, which is what the sensitivity table is for. Seventh, if the trial concludes “no difference worth acting on”, that is the trial’s conclusion, and the checklist in this book should be revised accordingly.

Ethics and conflicts of interest

The trial involves human participants and must be approved by the ethics committee of the implementing institution. The risk is minimal: participants only find buttons on fictional pages for about 15 minutes; payment follows the participant pool’s standard; only age band, device type and task data are collected, with no names, contact details or other personally identifying information; participants may withdraw at any time without consequence. The protocol, pages, code and de-identified data will be published, and the results will be published whether positive or negative. To limit the influence of interests, the author takes no part in recruitment, data collection or analysis, has no access to the allocation sequence, and the trial uses no page or data from the author’s companies or their clients.

Disclosure

The author uses the works of 5D Library, including this book, as the methodological basis of his community courses, the 5D Ascension Cards (Dimensional Shift Card Decks) and the Yafu AI knowledge base, and therefore has a commercial interest in the subject of this book. The book promotes no course, card deck or service; the Chapter 8 trial is designed to be run and analysed by researchers independent of the author’s business and does not recruit from the author’s community, students or customers; all the code and the study protocol are published openly for anyone to check. The author founded and runs Asia Influencer X and Yafu Global, which work in influencer marketing, live-stream commerce, digital marketing and website design and construction; this work involves the interface-design and conversion questions discussed in this book, so he has a direct commercial interest in its subject. The book uses no data from these two companies or their clients as evidence and does not promote their services.

Chapter summary

The four-step checklist examines an interface for salience, legibility, choice and pointing, and evidence; the example shows that a revision can pass the first three steps while the fourth stays flagged until tested. The pre-registered online randomised controlled trial, run by independent researchers on pages for 12 fictional brands, compares checklist-revised layouts with the originals; its single primary outcome is the log time to a correct click (analysis of covariance), the smallest effect of interest a 10 % shorter geometric-mean time, with 158 analysable and 190 recruited per group for 90 % power, a decision rule written as code and a pre-specified “lower accuracy” qualifier; it will be published whatever the result. It tests only one bundle of revisions on one kind of task and device; it cannot prove the book’s other models, nor carry over directly to conversion or sales.

References

Borm, George F., Jaap Fransen, and Wim A. J. G. Lemmens. 2007. “A Simple Sample Size Formula for Analysis of Covariance in Randomized Clinical Trials.” Journal of Clinical Epidemiology 60 (12): 1234–38.
Brooke, John. 1996. “SUS: A ‘Quick and Dirty’ Usability Scale.” In Usability Evaluation in Industry, edited by Patrick W. Jordan, B. Thomas, Bernard A. Weerdmeester, and Ian L. McClelland. Taylor & Francis.
Nosek, Brian A., Charles R. Ebersole, Alexander C. DeHaven, and David T. Mellor. 2018. “The Preregistration Revolution.” Proceedings of the National Academy of Sciences 115 (11): 2600–2606. https://doi.org/10.1073/pnas.1708274114.

Fan, Y. (2026). Computational Aesthetics, Visual Saliency Maps and Cognitive Ergonomics in User Interface Design: From the Hick–Hyman Law and Spatial Frequency Channels to Saccadic Trajectory Modelling, Read Beside the Kaogongji and Classical Treatise on Architectural Methods [计算美学、视觉显著性图谱与人机界面认知工程] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23836258 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).