⏱️ Slide Timing: 0.5 min - Short video first, then we unpack it - Goal: a checklist you can run on any chart — in the news, at work, or in your own reports

⏱️ Slide Timing: 5 min (video ~4–5 min) - Click "Play on TED" — opens in the browser; go full screen - Ask the class to jot down each trick as it appears - After the video: quick round — "which trick surprised you most?"

⏱️ Slide Timing: 2 min - We trust pictures more than tables — our eyes do the maths for us, and that is the weakness - The video's point: the problem is not graphs, it is how they are built - Most misleading charts you meet at work are accidents, not lies — but the damage is the same

⏱️ Slide Timing: 3 min - The video's example: a truck ad where the axis started near 95%, so one brand looked far ahead - Ask: on the left, how much more reliable does Brand A look? (2–3×). Real gap: ~2 points - Rule: bar and column charts start at zero. Line charts may zoom in — but say so clearly

⏱️ Slide Timing: 2 min - The video lists both axes: y-axis not starting at zero, x-axis skipping intervals, scales left out - Uneven x-axis makes growth look smoother or sharper than it was - Quick habit: read the axis labels before you look at the shape

⏱️ Slide Timing: 3 min - An eye-catching unit (stacks of cash, football fields) feels meaningful but rarely helps a decision - 3D pies: slices at the front look bigger than the same slice at the back - Dual-axis charts: you can make almost any two lines cross by choosing the axis ranges

⏱️ Slide Timing: 3 min - The video's example: a weather chart that compared a dry spell only with the last few years, so it looked like near-record — older years had a longer streak - That one was probably careless, not dishonest — someone just picked a short date range - Ask of every time chart: why does it start *here*? What happens if I go further back?

⏱️ Slide Timing: 3 min - The flip side of the truncated axis: a huge range flattens a small change that is still important - Here, under one degree looks like nothing on the left — but for climate it is a big shift - Ignored variables: "crime went up" — did population go up too? Show per-capita - So "always start at zero" is not the whole rule; the right scale depends on what change matters

⏱️ Slide Timing: 2 min - Source check is the cheapest and fastest test - At work: whose dashboard is it, and what are they measured on? - Before forwarding a chart: can you name the original data source?

⏱️ Slide Timing: 2 min - Turn the lens around: the most common misleading chart you will see this year might be your own - Excel / Power BI pick axis ranges automatically — and often not from zero

⏱️ Slide Timing: 3 min - Source: Mahbub et al., "The Perils of Chart Deception", IEEE VIS 2025 — 800 chart pairs, 8 trick types, 16,000+ answers - Models were 2025 versions (GPT-4o, Gemini 2.5 Pro, Claude 3.7 Sonnet, plus open-source models) — newer ones may do better, but the lesson holds - Gemini 2.5 Pro was the most robust: fooled by 3 of 8 trick types; spotted the upside-down axis in 79% of cases - None of the open-source models mentioned the upside-down axis in their answers - Upside-down axis: higher values plotted lower, so a rise looks like a fall — not covered in the video, but worth adding to the checklist - Ask: who has pasted a chart into ChatGPT / Claude and asked "what does this show"?

⏱️ Slide Timing: 2 min - Seven quick questions — takes under a minute per chart - Pair exercise: find one chart in today's news and run the checklist together

⏱️ Slide Timing: 1 min - If they remember one thing: read the axes before the shape

⏱️ Slide Timing: 1 min - The TED-Ed lesson page has a short quiz and discussion questions — good homework - Huff is old but still the most readable book on the topic

⏱️ Slide Timing: 3 min - Ask: "Name a chart you've seen at work that failed one of the checks" - Link back to the Dashboard Design deck: same rules apply when you build your own reports