-
Notifications
You must be signed in to change notification settings - Fork 0
/
algorithmic-bias.html
432 lines (372 loc) · 17.3 KB
/
algorithmic-bias.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
<!doctype html>
<html lang="en">
<head>
<title>LIRNEasia · Visualizations</title>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<meta name="author" content="Viren Dias">
<link rel="stylesheet" href="https://fonts.googleapis.com/icon?family=Material+Icons">
<link rel="stylesheet" href="https://unpkg.com/material-components-web@latest/dist/material-components-web.min.css">
<link rel="stylesheet" href="./css/app.css">
</head>
<body class="mdc-typography">
<header class="mdc-top-app-bar mdc-top-app-bar--short mdc-top-app-bar--short-collapsed mdc-top-app-bar--short-has-action-item" data-mdc-auto-init="MDCTopAppBar" style="top: 0px;">
<div class="mdc-top-app-bar__row">
<section class="mdc-top-app-bar__section mdc-top-app-bar__section--align-start">
<a href="./" class="mdc-icon-button material-icons mdc-top-app-bar__navigation-icon">home</a>
<section class="mdc-top-app-bar__section mdc-top-app-bar__section--align-end">
<a href="https://lirneasia.net" target="_blank" class="mdc-button mdc-button--unelevated" data-mdc-auto-init="MDCRipple">
<div class="mdc-button__ripple"></div>
<span class="mdc-button__label">LIRNEasia</span>
</a>
</section>
</div>
</header>
<main>
<h1 class="mdc-typography--headline2">Algorithmic Bias</h1>
<div class="container">
<div class="mdc-chip-set mdc-chip-set--choice" data-mdc-auto-init="MDCChipSet" role="grid">
<div class="chip-center">
<div class="mdc-chip mdc-chip--selected" d="group">
<div class="mdc-chip__ripple"></div>
<span role="gridcell">
<span role="button" tabindex="0" class="mdc-chip__primary-action">
<span class="mdc-chip__text">Risk Score Group</span>
</span>
</span>
</div>
<div class="mdc-chip" d="positive">
<div class="mdc-chip__ripple"></div>
<span role="gridcell">
<span role="button" tabindex="1" class="mdc-chip__primary-action">
<span class="mdc-chip__text">Recidivants</span>
</span>
</span>
</div>
<div class="mdc-chip" d="negative">
<div class="mdc-chip__ripple"></div>
<span role="gridcell">
<span role="button" tabindex="2" class="mdc-chip__primary-action">
<span class="mdc-chip__text">Non-Recidivants</span>
</span>
</span>
</div>
</div>
</div>
<div id="algorithmic-bias-vis"></div>
<p class="mdc-typography--body1">
This is a series of bar charts depicting the results of the COMPAS algorithm in the US—which assigns risk scores
based on how likely a defendant will recidivate—against the actual two-year outcome. Each bar chart depicts the
same data, but from a different perspective. Click or tap on any of the chips above the chart to change
perspective. Notice how the algorithm appears to be fair with regard to ethnicity in some perspectives but not
others.
</p>
</div>
</main>
<script src="https://unpkg.com/material-components-web@latest/dist/material-components-web.min.js"></script>
<script src="https://unpkg.com/@popperjs/core@2"></script>
<script src="https://unpkg.com/tippy.js@6"></script>
<script src="https://d3js.org/d3.v5.min.js"></script>
<script src="https://unpkg.com/topojson@3"></script>
<script>
window.mdc.autoInit();
// Render the visualization ----------------------------------------------------------------------------------------
d3.csv("./resources/recidivism_bias.csv")
.then(function(biasData) {
// Initialize the canvas ---------------------------------------------------------------------------------------
const width = 900;
const height = 600;
const topPadding = 50;
const rightPadding = 5;
const bottomPadding = 70;
const leftPadding = 100;
const plotPadding = 5;
d3.select("#algorithmic-bias-vis")
.append("svg")
.attr("preserveAspectRatio", "xMidYMin meet")
.attr("viewBox", [0, 0, width, height])
.style("max-height", "100vh")
.style("max-width", "100vw")
.append("g")
const svg = d3.select("#algorithmic-bias-vis")
.select("svg")
const chartGroup = d3.select("#algorithmic-bias-vis")
.select("svg")
.select("g")
let svgZoom = svg.node().getBoundingClientRect().width / width;
// Parse the data ----------------------------------------------------------------------------------------------
let chartData = [];
for (var score = 1; score <= 10; score++) {
const blackGroup = biasData.filter(d => d.score == score)
.filter(d => d.ethnicity == "black")
.filter(d => d.recidivated == 1)
.length /
biasData.filter(d => d.score == score)
.filter(d => d.ethnicity == "black")
.length;
const blackPositive = biasData.filter(d => d.recidivated == 1)
.filter(d => d.ethnicity == "black")
.filter(d => d.score == score)
.length /
biasData.filter(d => d.recidivated == 1)
.filter(d => d.ethnicity == "black")
.length;
const blackNegative = biasData.filter(d => d.recidivated == 0)
.filter(d => d.ethnicity == "black")
.filter(d => d.score == score)
.length /
biasData.filter(d => d.recidivated == 0)
.filter(d => d.ethnicity == "black")
.length;
const whiteGroup = biasData.filter(d => d.score == score)
.filter(d => d.ethnicity == "white")
.filter(d => d.recidivated == 1)
.length /
biasData.filter(d => d.score == score)
.filter(d => d.ethnicity == "white")
.length;
const whitePositive = biasData.filter(d => d.recidivated == 1)
.filter(d => d.ethnicity == "white")
.filter(d => d.score == score)
.length /
biasData.filter(d => d.recidivated == 1)
.filter(d => d.ethnicity == "white")
.length;
const whiteNegative = biasData.filter(d => d.recidivated == 0)
.filter(d => d.ethnicity == "white")
.filter(d => d.score == score)
.length /
biasData.filter(d => d.recidivated == 0)
.filter(d => d.ethnicity == "white")
.length
chartData.push({
score: score,
black: {
group: blackGroup,
positive: blackPositive,
negative: blackNegative
},
white: {
group: whiteGroup,
positive: whitePositive,
negative: whiteNegative
}
});
}
// Render the chart area ---------------------------------------------------------------------------------------
const yScale = d3.scaleLinear()
.domain([0, 1])
.rangeRound([height - bottomPadding, topPadding]);
const xScale = d3.scaleBand()
.domain(chartData.map(d => d.score))
.rangeRound([leftPadding, width - rightPadding])
.padding(0.2);
const groupScale = d3.scaleBand()
.domain(["black", "white"])
.rangeRound([0, xScale.bandwidth()])
.padding(0.1);
const scoreGroup = chartGroup.selectAll("g")
.data(chartData)
.enter()
.append("g")
.attr("transform", d => "translate(" + xScale(d.score) + ",0)");
scoreGroup.append("rect")
.attr("class", "blacks")
.attr("x", groupScale("black"))
.attr("y", d => yScale(0))
.attr("width", groupScale.bandwidth())
.attr("height", 0)
.attr("fill", d3.schemeCategory10[0]);
scoreGroup.append("rect")
.attr("class", "whites")
.attr("x", groupScale("white"))
.attr("y", d => yScale(0))
.attr("width", groupScale.bandwidth())
.attr("height", 0)
.attr("fill", d3.schemeCategory10[1]);
// Render the axes ---------------------------------------------------------------------------------------------
const axisTitleFontSize = 14;
const axisFontSize = 12;
const axisTickLength = 10;
const xAxis = svg.append("g")
.attr("class", "axis")
.attr("transform", "translate(0, " + (height - bottomPadding + plotPadding) + ")")
.call(d3.axisBottom(xScale).tickSize(5))
xAxis.selectAll("text")
.style("color", "var(--mdc-theme-text-primary-on-background)")
.style("font-size", axisFontSize / (Math.pow(svgZoom, 0.7)))
xAxis.selectAll("path, line")
.style("color", "var(--mdc-theme-text-primary-on-background)")
.style("stroke-width", 1);
svg.append("text")
.text("COMPAS risk score")
.attr("class", "mdc-typography--body1 axis-title")
.attr("x", leftPadding + (width - leftPadding - rightPadding) / 2)
.attr("y", height - bottomPadding + plotPadding + 30)
.style("text-anchor", "middle")
.style("alignment-baseline", "text-before-edge")
.style("fill", "var(--mdc-theme-text-primary-on-background)")
.style("font-size", axisTitleFontSize / (Math.pow(svgZoom, 0.7)));
const yAxis = svg.append("g")
.attr("class", "axis")
.attr("transform", "translate(" + (leftPadding - plotPadding) + ", 0)")
.call(d3.axisLeft(yScale).tickFormat(d3.format(".0%")).tickSize(5));
yAxis.selectAll("text")
.style("color", "var(--mdc-theme-text-primary-on-background)")
.style("font-size", axisFontSize / (Math.pow(svgZoom, 0.7)))
yAxis.selectAll("path, line")
.style("color", "var(--mdc-theme-text-primary-on-background)")
.style("stroke-width", 1);
const yLabel = svg.append("text")
.text("Percentage of defendants that recidivated in each group")
.attr("class", "axis-title mdc-typography--body1")
.attr(
"transform",
"translate(" + (leftPadding - 60) + "," + (topPadding + (height - topPadding - bottomPadding) / 2) +
") rotate(-90)"
)
.style("text-anchor", "middle")
.style("alignment-baseline", "text-after-edge")
.style("fill", "var(--mdc-theme-text-primary-on-background)")
.style("font-size", axisTitleFontSize / (Math.pow(svgZoom, 0.7)));
// Render the legend -------------------------------------------------------------------------------------------
const legendFontSize = 14;
const legendItemSpacing = 120;
const dimension = 20
const legendGroup = svg.append("g")
.attr("transform", "translate(" + leftPadding + ", 10)");
legendGroup.append("rect")
.attr("x", 0)
.attr("y", 0)
.attr("width", dimension)
.attr("height", dimension)
.style("fill", d3.schemeCategory10[0]);
legendGroup.append("text")
.text("Blacks")
.attr("class", "legend mdc-typography--body1")
.attr("x", dimension + 4)
.attr("y", dimension / 2)
.style("text-anchor", "start")
.style("alignment-baseline", "central")
.style("fill", "var(--mdc-theme-text-primary-on-background)")
.style("font-size", legendFontSize / (Math.pow(svgZoom, 0.7)));
legendGroup.append("rect")
.attr("x", legendItemSpacing)
.attr("y", 0)
.attr("width", dimension)
.attr("height", dimension)
.style("fill", d3.schemeCategory10[1]);
legendGroup.append("text")
.text("Whites")
.attr("class", "legend mdc-typography--body1")
.attr("x", legendItemSpacing + dimension + 4)
.attr("y", dimension / 2)
.style("text-anchor", "start")
.style("alignment-baseline", "central")
.style("fill", "var(--mdc-theme-text-primary-on-background)")
.style("font-size", legendFontSize / (Math.pow(svgZoom, 0.7)));
// Add mouse hover behaviour for each bar ----------------------------------------------------------------------
chartGroup.selectAll("g")
.on("mouseover", function() {
d3.select(this)
.select(".blacks")
.style("fill", d3.color(d3.schemeCategory10[0]).brighter(0.5));
d3.select(this)
.select(".whites")
.style("fill", d3.color(d3.schemeCategory10[1]).brighter(0.5));
})
.on("mouseout", function() {
d3.select(this)
.select(".blacks")
.style("fill", d3.schemeCategory10[0]);
d3.select(this)
.select(".whites")
.style("fill", d3.schemeCategory10[1]);
})
// Add popovers for each bar group -----------------------------------------------------------------------------
chartGroup.selectAll("g")
.attr("popover", true)
const popoverInstances = tippy('[popover=true]', {
content: "placeholder",
allowHTML: true,
theme: "custom",
maxWidth: "none",
arrow: false,
placement: "top",
});
popoverInstances.forEach(popover => {
popover.popper.querySelector(".tippy-box").classList.add("mdc-elevation--z1");
});
// Update the chart --------------------------------------------------------------------------------------------
var changePerspective = function(perspective) {
const maxValue = d3.max(
chartData.map(d => d.black[perspective])
.concat(chartData.map(d => d.white[perspective]))
);
yScale.domain([0, maxValue]);
yAxis.transition()
.duration(750)
.call(d3.axisLeft(yScale).tickFormat(d3.format(".0%")).tickSize(5));
yAxis.selectAll("text")
.style("color", "var(--mdc-theme-text-primary-on-background)")
.style("font-size", axisFontSize / (Math.pow(svgZoom, 0.7)))
yAxis.selectAll("path, line")
.style("color", "var(--mdc-theme-text-primary-on-background)")
.style("stroke-width", 1);
d3.selectAll(".blacks")
.transition()
.duration(750)
.attr("x", groupScale("black"))
.attr("y", d => yScale(d.black[perspective]))
.attr("width", groupScale.bandwidth())
.attr("height", d => yScale(0) - yScale(d.black[perspective]));
d3.selectAll(".whites")
.transition()
.duration(750)
.attr("x", groupScale("white"))
.attr("y", d => yScale(d.white[perspective]))
.attr("width", groupScale.bandwidth())
.attr("height", d => yScale(0) - yScale(d.white[perspective]));
popoverInstances.forEach(popover => popover.setContent(element => {
const eleData = d3.select(element).data()[0];
const contentHead = "<h2 class='mdc-typography--headline6'>Risk Score: " + eleData.score + "</h2>";
const contentBody = "<p class='mdc-typography--body2'>" +
"<i class='material-icons' style='vertical-align: middle; color:" + d3.color(d3.schemeCategory10[0]) +
";'>stop</i>" + d3.format(".1%")(eleData.black[perspective]) + " blacks</p>" +
"<p class='mdc-typography--body2'>" + "<i class='material-icons' style='vertical-align: middle; color:" +
d3.color(d3.schemeCategory10[1]) + ";'>stop</i>" + d3.format(".1%")(eleData.white[perspective]) +
" whites</p>";
return contentHead + "<div class='popover-divider'></div>" + contentBody;
}));
if (perspective == "group") {
yLabel.text("Defendants recidivated")
} else if (perspective == "positive") {
yLabel.text("Recidivants assigned")
} else if (perspective == "negative") {
yLabel.text("Non-recidivants assigned")
}
}
changePerspective("group");
window.onresize = function() {
svgZoom = svg.node().getBoundingClientRect().width / width;
document.querySelectorAll("svg .axis text").forEach(element => {
d3.select(element).style("font-size", axisFontSize / (Math.pow(svgZoom, 0.7)));
});
document.querySelectorAll("svg .axis-title").forEach(element => {
d3.select(element).style("font-size", axisTitleFontSize / (Math.pow(svgZoom, 0.7)));
});
document.querySelectorAll("svg .legend").forEach(element => {
d3.select(element).style("font-size", legendFontSize / (Math.pow(svgZoom, 0.7)));
});
}
// Define button behaviour -------------------------------------------------------------------------------------
d3.selectAll(".mdc-chip")
.on("click", function() {
d3.selectAll(".mdc-chip").classed("mdc-chip--selected", false);
d3.select(this).classed("mdc-chip--selected", true);
changePerspective(d3.select(".mdc-chip--selected").attr("d"));
});
})
</script>
</body>
</html>