5D Library › Computational Aesthetics, Visual Saliency Maps and Cognitive Ergonomics in User Interface Design · 中文
Preface: Seeing, Choosing and Clicking
Note
Scope: please read this first. This book presents mathematical models and a review of evidence for teaching and research. It is not a design standard, nor business, marketing, legal or accessibility-compliance advice, and it promises no gain in conversion, click-through or sales. Every web page, button, brand, user and experimental figure in the book is fictional or simulated, not data from any real website, company or person. The models hold only under the assumptions stated; real users’ behaviour depends on task, experience, culture, device and setting, and can violate those assumptions at any time. The legal standing of accessibility requirements such as WCAG differs between countries and regions; for compliance questions, consult a professional. The book offers no method that “guarantees”, “always wins” or “raises conversion by X %”.
A visitor opens a product page. Within a second his eyes have already jumped two or three times: first to a large photograph, then to the logo at the top left, and only then to the button that says “Add to basket”. He must decide which of five or six options to take and then move the pointer there and click. Designers often organise this process with a few simple rules: divide the page by the golden ratio 1.618, make every piece of content reachable within three clicks, and make the button big, near and strongly contrasting. This book asks which of these rules are supported by mathematics and evidence, which need qualifying, and which are wrong.
Answering takes four layers of tools. The first is choice and action: how long it takes a person to decide among several options and then move a hand to the target, and whether the two times can be optimised together. The second is early vision: how, before anyone consciously “looks”, the visual system splits a scene into channels of spatial frequency, colour and orientation, and which locations therefore “pop out” first. The third is eye movements: how the point of gaze jumps across a page, how much of this a model can predict, and where the unpredicted part comes from. The fourth is beauty and complexity: whether a layout “looking good” can be measured, and what such a measure would measure.
What the book covers
Chapter 1 checks two popular claims, the golden ratio and the “three-click rule”; Chapter 2 covers the Hick–Hyman law and Fitts’s law and a model of the time taken by the whole “search–decide–point” chain; Chapter 3 covers visual search and the Itti–Koch–Niebur saliency model; Chapter 4 covers eye movements, centre bias, saccadic trajectories and inhibition of return; Chapter 5 covers spatial-frequency channels, the contrast sensitivity function and the relative luminance and contrast ratio of WCAG; Chapter 6 covers type scales, Birkhoff’s aesthetic measure, compression complexity and the evidence on first impressions; Chapter 7 reads these beside the Kaogongji of the Rites of Zhou and the modular cai–fen system of Li Jie’s Yingzao fashi; and Chapter 8 offers a four-step checklist and a pre-registered online randomised controlled trial.
The book neighbours several others in 5D Library; where they overlap, it only recaps briefly, cites the earlier book and does not repeat its programs. Book 3, Sacred Cubit Geometry, treated length standards and geometric proportion. Book 6, The Economics of Attention, treated the scarcity of attention and conversion funnels. Book 24, Attention Networks, Working-Memory Bottlenecks and Self-Determination Theory, treated Posner’s orienting network and working-memory capacity (Chapters 3 and 4 only recap). Book 31, Architectural Acoustics, Reverberation Time and the Physical Models of Environmental Psychology, treated fractal visual dimension and the “Master Builder” section of the Kaogongji (Chapters 6 and 7 only recap). Book 33, Scientific Methodology, Falsifiability and the Mathematical Statistics of the Replication Crisis, treated pre-registration, power and decision rules (Chapters 1 and 8 follow it). Book 34, Psychophysics, the Weber–Fechner Law and Bayesian Multisensory Integration, treated Weber’s law, the extra assumption Fechner needed and Stevens’s power law (Chapters 5 and 6 only recap). Book 43, Embodied Cognition, Bernstein’s Degrees-of-Freedom Problem and the Control Theory of Motor Synergies, treated the difference between Fitts’s original and Shannon forms and the optimised-submovement model (Chapter 2 only recaps). The risk notice, disclosure and Chapter 8 of this book follow the pattern of Book 45.
Claims corrected after checking
Several points in the author’s framework needed correcting, supplementing or qualifying after checking. Each chapter says so where it arises, and the accompanying editorial notes list them together.
- The “universal myth” of the golden ratio. The claim that the framework sets out to refute does indeed lack support. Rectangle-preference experiments since Fechner (1876) give weak, method-dependent results: with paired comparisons the group preference for the golden ratio is weak, while individual preferences are stable and differ widely (McManus 1980; Green 1995; Spehar et al. 2016); the instructions given and the participants’ art training alone can move the average preference from the golden ratio to the square (Hekkert et al. 1994). This round of checking found no peer-reviewed study showing that 1.618 raises web conversion. The program also shows that a layout with only 8 boxes contains, with probability 0.795, some ratio within 1 % of 1.618 by chance, so “this successful page hides a golden ratio” is no evidence (Chapter 1).
- The “three-click rule”. Porter’s 2003 study is industry research published by User Interface Engineering (UIE), not a peer-reviewed paper: across 44 users, 620 tasks and more than 8,000 clicks, the number of clicks was unrelated to task completion, and satisfaction did not vary with it either (Porter 2003); the Nielsen Norman Group has published an article to the same effect (Laubheimer 2019). Peer-reviewed studies of menu depth and breadth conclude that “depth has a cost, but information scent and the shape of the hierarchy matter as much”, not “three” (Miller 1981; Kiger 1984; Larson and Czerwinski 1998). The word “unconditional” in the framework is the right target of criticism; the book writes “the number of clicks is not a valid design constraint”, not “the number of clicks does not matter” (Chapter 1).
- The Hick–Hyman formula. The framework writes . Hick (1952) fitted choice reaction times with equally likely alternatives by , where the stands for the extra alternative of “no signal” (Hick 1952; Proctor and Schneider 2018); Hyman (1953) used the information entropy , with no (Hyman 1953). The framework’s expression mixes the two: with equal probabilities it equals exactly (the program checks this term by term), and with unequal probabilities it is neither Hick’s nor Hyman’s form, staying near 1 bit when one option is almost certain while the entropy goes to 0. The book uses Hyman’s form and explains the origin of the (Chapter 2).
- “Minimising the total time of the whole chain”. Adding the times for visual search, decision and pointing gives an optimisation problem that can be solved only after the layout variables, the distribution of click probabilities and the kind of search (serial or parallel, novice or expert) are stated, and the solution changes with those assumptions (Cockburn et al. 2007). With the logarithmic form of Fitts’s law, the optimal button widths are proportional to the click probabilities and the expected pointing difficulty equals a constant plus Hyman’s entropy; for novices, serial reading makes “more options, slower” roughly linear, not logarithmic. The framework’s “minimisation equation” is rewritten as “optimisation under stated assumptions” (Chapter 2).
- The Itti–Koch–Niebur model. Nine-level dyadic Gaussian pyramids, centre , surround , , 42 feature maps, the normalisation operator and winner-take-all with inhibition of return all check out against the paper (Itti et al. 1998). But the paper itself states that the model has no top-down component and that its attentional trajectories are “not necessarily” similar to human fixations; on web pages, centre bias, a top-left bias, text and faces often predict fixations better than low-level saliency (Tatler 2007; Borji and Itti 2013; Shen and Zhao 2014). The book’s implementation works on a reduced 256 × 192 synthetic page and replaces Gabor filters with second differences; every difference is stated (Chapters 3 and 4).
- WCAG relative luminance. The sRGB linearisation threshold is 0.03928 in the original WCAG 2.0 text and 0.04045 in IEC 61966-2-1; WCAG switched to the latter in May 2021, noting that the change has no practical effect on the calculations (W3C 2023; IEC 1999). The program proves it: no 8-bit integer value lies between the two thresholds (10/255 = 0.03922, 11/255 = 0.04314), so the results for 8-bit colour are identical. The framework does not mention APCA; it should be said that APCA was removed from the WCAG 3 working drafts in 2023 and that the contrast algorithm of WCAG 3 has still not been decided (Roselli 2026) (Chapter 5).
- The “Weber–Fechner law of type scales”. Modular type scales in design (1.2, 1.25, 1.333 and so on) are design conventions, not a law derived from a Weber fraction; Weber’s law says only that size differences above some ratio can be told apart and favours no particular ratio. Book 34 has discussed the extra assumption needed to go from Weber to Fechner. The book presents the “type ladder” as the author’s model, with assumed Weber fractions (Chapter 6).
- Birkhoff and Kolmogorov. Birkhoff’s 1933 aesthetic measure is reproduced (Birkhoff 1933), but it gives every rectangle of ordinary (not too elongated) proportions the same score of 1.25 and so cannot single out the golden rectangle; the square scores higher (1.50). Kolmogorov complexity is uncomputable and can be represented only by compressed length as an upper bound (Li and Vitányi 2008). The evidence on visual complexity and first impressions (the 50-millisecond study of Lindgaard and colleagues, 2006; Tuch and colleagues, 2012; Reinecke and colleagues, 2013) is reported with its samples and effects (Lindgaard et al. 2006; Tuch et al. 2012; Reinecke et al. 2013) (Chapter 6).
- The classical texts and building module. Kaogongji: “审曲面埶,以饬五材,以辨民器,谓之百工”. The online transcription available in this round (Chinese Text Project) reads 势; the author’s framework and the traditional commentary editions read 埶, which is read as 势. The paper copy of the edition the author specified, Sun Yirang’s Zhouli zhengyi (Zhonghua Shuju), could not be consulted in this round; the book provisionally follows 埶, notes its reading, and asks the author to settle the character against that edition (孙诒让 撰,王文锦、陈玉霞 点校 1987). Yingzao fashi, juan 4: “凡构屋之制,皆以材为祖;材有八等,度屋之大小,因而用之” and “各以其材之广,分为十五分,以十分为其厚” check out (online transcriptions punctuate the first clause with either a full stop or a semicolon); all eight grades of cai have a depth-to-breadth ratio of 15:10 = 3:2, 7.3 % from the golden ratio, and the “full cai” (zucai) is 21:10 (李诫 撰;梁思成 注释 1983). The comparison of the human visual field with screen proportions is an analogy only (Chapter 7).
- Chapter 8’s outcome. The framework proposes comparing an “optimised layout” with a “control layout”. The book sets a single primary outcome, the log time to a correct click on the target; first-click accuracy is a secondary outcome, for reasons given in Chapter 8 (Chapter 8).
How to read the boxes
As in the other books of 5D Library, different kinds of statement are marked separately.
Evidence
Evidence boxes report published research findings, with sources.
Calculation
Calculation boxes give short Python programs. Each runs unchanged in a standard Python 3 environment, uses only the standard library and fixes its random seed; its actual output is printed below it. All pages, buttons, users and experimental data in the book are fictional or synthetic, and parameters are illustrative, not estimates for any real website or population.
Hypothesis · not established
Hypothesis boxes contain the author’s ideas, marked “not established”, together with the result that would show them wrong. The author’s own models and definitions are labelled “the author’s model” in the text.
Source text
Source boxes quote ancient texts with a plain-language rendering.
Note
Note boxes give reminders and explain choices.
About this edition
This is version 1.0.0, the first complete edition: a preface and eight chapters, in Chinese and English. It contains 16 Python programs and 5 vector figures. The text, boxes and code were drafted with an AI writing tool from the author’s title, framework and notes; the author is responsible for reviewing and revising them and for checking sources. AI-drafted text can get bibliographic details wrong; corrections are welcome through the 5D Library website. The menu-tree model of Chapter 1, the toolbar chain model of Chapter 2, the observer-mixture and fixation-timing models of Chapter 4, the type-ladder model of Chapter 6 and the checklist limits of Chapter 8 are the author’s models; their mathematical properties are not evidence about real users.
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.
References
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).