Course content
Open a chapter to read its review, then practise with its question bank.
Foundations
-
Chapter 1 · Week 1
Visualization in the Age of AI
what visualization is and why it works, Anscombe's quartet and the Datasaurus, the visualization pipeline (data, transformation, visual mapping, view), external cognition, interaction and Shneiderman's mantra, a short history (Playfair, Snow, Nightingale, Minard), the Python visualization ecosystem (matplotlib, Altair/Vega-Lite, Plotly, Bokeh, Streamlit) and BI tools, what AI can and cannot do for visualization, the course's AI-use rules, the CityPulse case study
-
Chapter 2 · Week 2
Data Abstraction: Datasets, Attributes and Tidy Data
items and attributes, dataset types (tables, networks and trees, fields, geometry, sets and lists), attribute types (categorical, ordinal, quantitative) and ordering direction (sequential, diverging, cyclic), keys and values, temporal and spatial semantics, mapping attribute types to Altair's Q/N/O/T, tidy data (variables, observations, observational units), wide vs. long format (melt and pivot), file formats (CSV, JSON, Parquet, Excel), dtypes, missing and invalid values, nested JSON from APIs, data quality checks and AI-written cleaning code
-
Chapter 3 · Week 3
Task Abstraction: Why Is the User Looking?
tasks as {action, target} pairs, the action hierarchy (analyze: discover, present, enjoy, annotate, record, derive; search: lookup, locate, browse, explore; query: identify, compare, summarize), targets (trends, outliers, features, distribution, extremes, dependency, correlation, similarity, topology, paths, shape), writing tasks from stakeholder statements, chains of tasks and the views they require, derived attributes as design, the nested model of visualization design and validation (domain, abstraction, idiom, algorithm), threats and matching validation methods, AI-generated task analyses
-
Chapter 4 · Week 4
Getting Data: Files, Databases, APIs and Web Scraping
where visualization data comes from, relational databases and SQL (SELECT, WHERE, GROUP BY, HAVING, JOIN, LEFT JOIN), pandas with SQLite and MySQL (read_sql_query, to_sql), parameterized queries and SQL injection, REST APIs and JSON, HTTP status codes, pagination, rate limits and retries, API keys and secrets, web scraping with requests and BeautifulSoup, robust selectors, dynamic pages, robots.txt and the Robots Exclusion Protocol, terms of service, copyright and personal data, open-data standards (GBFS), reproducible acquisition pipelines
Visual Encoding
-
Chapter 5 · Week 5
Marks, Channels and Perception
marks (points, lines, areas; connection and containment), channels (position, length, angle, area, colour hue, luminance, saturation, shape, motion), identity vs. magnitude channels, expressiveness and effectiveness, channel rankings, Cleveland and McGill's graphical perception experiments and their crowdsourced replication, Stevens' power law, accuracy, discriminability, separability (integral vs. separable channels), pop-out, Gestalt grouping, Weber's law and aligned scales, bar charts and the zero baseline, Tufte's lie factor, one dataset with many encodings in Altair
-
Chapter 6 · Week 6
Colour, Accessibility and Honest Charts
colour as luminance, saturation and hue, perceptual colour spaces (CIELAB, HCL), categorical, sequential, diverging and cyclic colour maps, why rainbow maps mislead, perceptually uniform maps (viridis), ColorBrewer and colour-blind-safe palettes (Okabe–Ito), colour-vision deficiency and its simulation, WCAG contrast ratios and non-colour cues, alt text and accessible charts, misleading techniques (truncated axes, dual axes, cherry-picked windows, area and 3D, inconsistent scales, missing normalization), Tufte's data-ink and chartjunk, showing uncertainty, auditing AI-generated captions
-
Chapter 7 · Week 7
The Grammar of Graphics: Tables, Time and Distributions
the grammar of graphics (Wilkinson; ggplot2; Vega-Lite), Altair's building blocks (data, transforms, marks, encodings, scales, guides), transforms in the chart vs. in pandas (filter, aggregate, calculate, bin, timeUnit, window, fold), arranging tables by keys and values (scatter plots, bar and dot charts, heat maps), sorting and alignment, time-series idioms (line charts, aspect ratio, cycle and calendar heat maps), distribution idioms (histograms and bin width, box plots, violin and density plots, strip plots, ECDF), composition (layers, concatenation, facets, repeat), the 5,000-row limit and how to handle larger data, reviewing AI-generated Altair code
Interaction and Dashboards
-
Chapter 8 · Week 8
Interaction and Multiple Linked Views
why interaction (scale, multiple tasks, exploration), manipulating views (change, select, navigate: zoom, pan, semantic zoom), details on demand and tooltips, faceting (juxtapose, partition, superimpose), coordinated multiple views and brushing and linking, overview + detail and focus + context, Altair parameters and selections (point, interval, bound to legends and widgets, conditions, filters), Plotly Express interaction, Bokeh 3 (ColumnDataSource, glyphs, tools, widgets, CustomJS vs. server callbacks), level of detail and response-time limits, animation vs. small multiples, the cost of interaction
-
Chapter 9 · Week 10
Dashboards: Design, Development and Deployment
what a dashboard is (Few) and the variety of dashboard genres (operational, analytical, strategic, public), designing from tasks and audiences, KPIs with context (targets, previous periods, benchmarks), sparklines and bullet graphs, layout and visual hierarchy, information density, data freshness, dashboard frameworks (Streamlit, Dash, Panel, Bokeh server) and BI tools (Tableau, Power BI, Superset, Metabase, Grafana), the Streamlit execution model, caching (cache_data, cache_resource, TTL), session state, connecting to databases and APIs, secrets, testing dashboards (Streamlit AppTest), deployment and performance, AI copilots in BI
Complex Data
-
Chapter 10 · Week 11
Maps, Networks and Trees
geographic coordinates and great-circle distance, map projections (Mercator, Web Mercator, equal-area), GeoJSON and TopoJSON, thematic maps (choropleths, proportional symbols, dot density, heat maps, flow maps, cartograms), normalization and classification (equal interval, quantile, natural breaks), the modifiable areal unit problem, maps in Altair (geoshape) and other tools, networks (node-link, force-directed layouts, adjacency matrices, arc diagrams), network metrics (degree, strength, betweenness, density) with networkx, the hairball problem and edge bundling, trees (node-link, indented, treemaps, sunbursts, icicles) and treemap layouts
-
Chapter 11 · Week 12
Scale: Reduction, Aggregation and High-Dimensional Data
too many items and too many attributes, filtering and aggregation (with care for what is lost), overplotting and its remedies (transparency, jitter, smaller marks, 2D binning, hexbins, datashader), sampling and its risks for outliers, many attributes (scatterplot matrices, parallel coordinates, heat maps of tables), dimensionality reduction (PCA with standardization and loadings, MDS, t-SNE, UMAP) and how to read embeddings, clustering (k-means and the elbow) as aggregation, embedding views of AI data, honest reporting of reduced views
AI and Visualization
-
Chapter 12 · Week 13
AI for Visualization: From Natural Language to Charts
the landscape of AI-assisted visualization (chart recommendation, natural-language interfaces, code assistants, chart question answering, automatic captions and alt text), how NL2VIS systems work (data summary, intent and task inference, specification generation, rendering), ambiguity and clarification, Vega-Lite as a target language, guardrails (JSON and schema validation, field and type checks, linting, execution checks, human review), evaluating NL2VIS (exact match, component accuracy, execution accuracy, human judgment, task success), chart question answering and multimodal models, fact-checking generated captions, privacy (schema-only prompts) and prompt injection through data, provenance and disclosure
-
Chapter 13 · Week 13
Visualization for AI and Real-Time Systems
visualization across the machine-learning lifecycle (data, training, evaluation, deployment), confusion matrices, precision, recall and accuracy, ROC and precision–recall curves and choosing thresholds, calibration, feature importance (permutation, SHAP) and partial dependence, per-group metrics and fairness, explaining models to stakeholders, monitoring deployed models (data drift, PSI, prediction drift), real-time and streaming visualization (windows, rolling aggregates, percentiles), alerting with thresholds and hysteresis, service-level objectives and error budgets, observability stacks (Prometheus, Grafana), monitoring edge–cloud systems