Textbook of Usability

Introduction

This textbook has covered a wide range of material: from the millisecond-level processing of the perceptual system to the centuries-old proportions of Palladian architecture, from the controlled precision of Fitts's Law to the messy reality of clinical software usability. This final chapter steps back to ask a broader question: can design be a science? And if so, what kind of science is it? [Simon, 2019; Cross, 2007] The argument of this textbook is that usability is — or can be — an applied science: a discipline that draws on fundamental knowledge about human beings to produce designs that serve human needs effectively Vincenti, 1993. This chapter synthesises the evidence presented throughout the book and makes the case for a cumulative, evidence-based approach to design Paton, 2022.

The Evidence Hierarchy for Design

Not all design knowledge is created equal. The evidence supporting a design decision can range from personal intuition to replicated controlled experiments. A useful framework organises design evidence into levels, analogous to the evidence hierarchies used in medicine.

Level 1: Expert Opinion and Intuition

A designer's judgment, shaped by experience and training. Valuable but subject to individual bias, limited by personal experience, and difficult to replicate or verify.

Level 2: Evolved Practice

Design conventions that have survived a process of cultural selection over extended periods — architectural proportions, typographic conventions, navigation patterns Alexander, 1977. Chapter 9 argued that evolved practice represents accumulated evidence: designs that worked were imitated; designs that failed were abandoned. The evidence is observational and retrospective, but the sample size (centuries of practice) and the selection pressure (buildings must be lived in, books must be read) give it substantial weight.

Level 3: Design Heuristics and Expert Consensus

Principles such as Nielsen's 10 heuristics Nielsen, 1994, Shneiderman's golden rules Shneiderman, 1987, and Norman's design principles Norman, 2016. These represent distilled expert knowledge, informed by research but not derived from a single experiment. They occupy a middle ground between individual judgment and empirical evidence.

Level 4: Observational Studies

Usability tests, field studies, surveys, and analytics data that measure actual user performance and experience. These provide empirical evidence about specific designs in specific contexts. Their limitation is that they describe what happened, not necessarily why, and their results may not generalise beyond the specific conditions studied.

Level 5: Predictive Models

Fitts's Law Fitts, 1954, the KLM Card, 1983, cognitive load theory Sweller, 1988, and other models that predict human performance from first principles. These models are derived from controlled experiments, have been widely replicated, and can be applied to new designs without empirical testing. They provide the most generalisable form of design evidence, albeit for specific aspects of performance (speed, accuracy, memory load) rather than the full breadth of usability.

Level 6: Controlled Experiments

Randomised controlled experiments that isolate the effect of specific design features on specific outcomes. These provide the strongest evidence for causal claims ("Design A is faster than design B because of feature X"). Chapter 18 covered the methodology; the present chapter considers their role in the broader evidence landscape.

The Scientific Laws of Design

Throughout this textbook, we have encountered several quantitative laws that predict human performance:

Fitts's Law (Chapter 5)

MT = a + b × log2(D/W + 1) Predicts pointing time as a function of distance and target width Fitts, 1954. Validated across dozens of input devices and thousands of experiments MacKenzie, 1992. The most robust quantitative law in HCI.

Hick's Law (Chapter 4)

RT = a + b × log2(n + 1) Predicts choice reaction time as a function of the number of equally probable alternatives [Hick, 1952; Hyman, 1953]. Applies to decision among known options; does not apply to visual search for unknown targets.

Miller's Law (Chapter 3)

Working memory capacity ≈ 4 ± 1 chunks for novel information Cowan, 2001 Sets the limit on how many independent items a user can hold in working memory simultaneously. Cowan's revision Cowan, 2001 of Miller's original 7 ± 2 Miller, 1956 is the figure now taken as authoritative. The most important constraint on information display design.

The Power Law of Practice (Chapter 6)

Tn = T1 × n^(-α) Predicts how performance improves with practice. Applies to virtually all skill acquisition, from typing to menu navigation.

The Steering Law (Chapter 5)

T = a + b × (D/W) Predicts movement time through a constrained path (tunnel) Accot, 1997. Governs the difficulty of cascading menus and narrow drag paths.

Weber's Law (Chapter 2)

ΔI / I = constant The just-noticeable difference in a stimulus is a constant proportion of the stimulus intensity. Governs contrast sensitivity and the perception of differences.

The Complementarity of Evidence Sources

The strongest design arguments draw on multiple evidence sources that converge on the same conclusion.

Open Questions

Despite the substantial body of knowledge covered in this textbook, many important questions remain open.

Individual Differences

The laws and models in this textbook describe "average" human performance. But individual differences in perception, cognition, and motor skill are substantial. How should designs accommodate this variability? Adaptive interfaces that adjust to individual users offer one approach, but they introduce their own usability challenges (unpredictability, loss of consistency).

Cultural Variation

Most usability research has been conducted in Western, educated, industrialised, rich, democratic (WEIRD) populations. Design conventions that seem universal — reading left to right, navigating top to bottom, using red for danger — are culturally specific. How well do the principles in this textbook generalise across cultures?

Emotional and Aesthetic Dimensions

The laws and models covered here primarily address performance (speed, accuracy, cognitive load). But usability also encompasses satisfaction, trust, engagement, and delight — dimensions that are harder to quantify and predict. Vitruvius's venustas (beauty) is acknowledged but not yet well integrated into the predictive framework.

The Changing Nature of Interaction

The traditional model of usability — a user operating deterministic controls to accomplish defined tasks — is being disrupted by AI (Chapter 19) [Shneiderman, 2022; Norman, 2023]. Conversational interfaces, autonomous agents, and generative systems create interaction patterns for which existing models were not designed. New models, methods, and principles will be needed.

Sustainability and Ethics

Usability has traditionally focused on making designs work well for their users. But design also affects people who are not the direct users — communities affected by data collection, workers displaced by automation, environments degraded by electronic waste. A complete science of design must eventually account for these broader impacts.

A Framework for Evidence-Based Design

Drawing together the threads of this textbook, we can articulate a framework for evidence-based design:

1. Start with the science. Use the quantitative laws (Fitts's Law, Hick's Law, Miller's Law, cognitive load theory) to establish the fundamental constraints on the design. These laws identify what is physically and cognitively possible.
2. Apply design heuristics. Use Nielsen's heuristics, Shneiderman's golden rules, and Norman's principles to guide the thousands of design decisions that cannot each be individually modelled. Heuristics provide broad coverage and catch common patterns of failure.
3. Draw on evolved practice. Consult the design traditions of architecture, typography, and other mature disciplines for guidance on proportion, hierarchy, rhythm, and organisation. These evolved practices represent accumulated evidence from centuries of human experience.
4. Predict with models. Use GOMS, KLM, Fitts's Law calculations, and cognitive load estimation to make quantitative predictions about specific design alternatives. Models are most useful for comparing options and identifying bottlenecks.
5. Evaluate with experts. Conduct heuristic evaluations and cognitive walkthroughs to catch problems before involving users. Expert review is fast and inexpensive; use it early and often.
6. Test with users. Conduct usability testing to discover problems that models and experts miss. Observe real users performing real tasks. There is no substitute for this empirical grounding.
7. Measure at scale. Use A/B testing and analytics to measure the impact of design changes on real user behaviour at scale. Quantitative data from large populations provides the most definitive evidence of a design's effectiveness.
8. Iterate. No design is finished. Each round of evidence — from models, experts, users, and analytics — reveals problems and opportunities that feed the next iteration.

Conclusion

The designed world — our buildings, our software, our tools, our cities — mediates our relationship with each other and with the tasks of daily life. When this mediation works well, we accomplish our goals efficiently and comfortably. When it fails, we struggle, err, and sometimes come to harm. The science of usability offers a path to better design: not a guarantee of perfection, but a systematic approach to understanding human capabilities, predicting human performance, and evaluating designed artefacts against human needs. The laws, models, heuristics, and methods presented in this textbook are tools for that work. The tools are imperfect and incomplete. The laws cover specific aspects of performance but not the full richness of human experience. The models predict expert behaviour but not the full range of human variability. The heuristics provide guidance but not certainty. And the methods reveal problems but do not automatically solve them. What the science does provide is a foundation — a body of knowledge, accumulated through experiment and practice, that makes informed design decisions possible. The architect who knows the proportions of Palladio, the interface designer who knows Fitts's Law, the clinician who understands cognitive load — all are better equipped to create designs that serve human beings well. The ambition of a science of design is to make this knowledge cumulative: each study, each validated model, each refined heuristic adds to a growing body of evidence that future designers can draw upon. This textbook is a contribution to that accumulation — an attempt to gather, organise, and make accessible the scientific knowledge that can make designed objects more usable, more efficient, and more humane.