-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
291 lines (198 loc) · 8.34 KB
/
index.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
<!DOCTYPE html>
<html>
<head>
<title>be_ethical()</title>
<meta charset="utf-8">
<meta name="author" content="Doug Friedman" />
<meta name="date" content="2018-10-30" />
<link href="libs/remark-css/default.css" rel="stylesheet" />
<link href="libs/remark-css/default-fonts.css" rel="stylesheet" />
</head>
<body>
<textarea id="source">
class: center, middle, inverse, title-slide
# be_ethical()
## Some practical advice for recognizing and handling ethical issues in data science
### Doug Friedman
### 10/30/2018
---
# What this talk isn't about
.center[]
---
# What this talk is about
1. Talk about common ethical issues you'll likely encounter in data science
--
2. Discuss my tips for handling them
--
3. Provide resources and libraries for tackling them
---
# Who am I?
I worked at Booz Allen Hamilton for 5+ years working with various federal agencies on a variety of analytics problems including population modeling and the effects of climate change.
.center[
]
Currently, I've been leading Johnson & Johnson's data science team here in Rhode Island at the Providence Healthcare Technology Center for over a year.
.center[]
---
class: inverse, center, middle
# Recognizing ethical issues
---
# An Illustrative Example
---
# An Illustrative Example
> ...researchers attending the...talk raised concerns during the Q&A afterward. How could the team be sure the training data were not biased to begin with? What happens when someone is mislabeled as a gang member? Lemoine asked rhetorically whether the researchers were also developing algorithms that would help heavily patrolled communities predict police raids.
--
> Hau Chan, a computer scientist now at Harvard University who was presenting the work, responded that he couldn’t be sure how the new tool would be used. “I’m just an engineer,” he said. Lemoine quoted a lyric from a song about the wartime rocket scientist Wernher von Braun, in a heavy German accent: “Once the rockets are up, who cares where they come down?” Then he angrily walked out.
--
---
# Common Ethical Issues
You don't fully understand **the context** for the problem you're solving.
--
- *Did these researchers look at the history of predictive policing?*
--
Your users don't **reasonably understand** the limitations of the solution you built.
--
- *Do these law enforcement officers understand what neural networks or black-box models can or can't do?*
--
You and/or your users **didn't think** about the potential ethical issues first.
--
- *Did these researchers and law enforcement officers discuss ethical issues in police work before development of this model?*
--
### This isn't limited to Machine Learning or AI! Any analysis or program can encounter these issues.
---
# Diving Deeper into Ethical Issues
Google [has a fantastic online course](https://developers.google.com/machine-learning/crash-course/fairness/video-lecture) that teaches you about a wide variety of human biases and how to recognize them in a machine learning context.
---
class: inverse, center, middle
# Tips for handling ethical issues
---
# Don't work alone!
--
.pull-left[
]
.pull-right[
* Find someone else to talk through your problem with...
- a friend
- a colleague
- a professor
* Ethical issues are complicated and a different perspective often shines a light on something you may have missed.
* Make your approach as transparent as possible so others can collaborate or review your work.
]
---
# Learn about your problem's context!
--
.pull-left[
* The issues you'll be asked to tackle with data science aren't new ones.
* Spend some time learning about the field you are jumping into and where people went wrong beforehand.
* A little research can go a long way to avoiding a serious issue later on.
]
.pull-right[
]
---
# Think about the ethics first!
--
.pull-left[
]
.pull-right[
* Map out your problem and think about where things might go wrong or affect others.
* Pick metrics that will help you track/measure these potential issues. _Hint: Default metrics often are **not** the best ones to track these._
* You can't expect to avoid ethical issues after you've already developed a solution.
]
---
# Be clear with your users!
--
* Not every user has a PhD in Computer Science!
* Make it easy for them to understand when and how to apply your solution.
* Unintended ethical issues can be more prevalent than intended ones.
.center[]
---
class: inverse, center, middle
# Some helpful resources
---
# IBM AI Fairness 360
.center[]
An [open-source Python toolkit](http://aif360.mybluemix.net/) that contains about a dozen different metrics/algorithms to aid in bias detection and avoidance for machine learning projects.
---
# Model Cards
.center[]
An [extensible, proven format](https://arxiv.org/abs/1810.03993) for sharing model information with your users.
.center[]
It even has an ethical considerations section!
---
# Interpretable Machine Learning
.pull-left[
]
.pull-right[
A [free textbook](https://christophm.github.io/interpretable-ml-book/) which provides step-by-step instructions on how to inspect and interpret all types of machine learning models.
* Partial Dependence Plots
* LIME (Locally Interpretable Model-Agnostic Explanations)
* Shapley Values
* ...and more!
]
---
# More Resources
- [Analyzing & Preventing Unconscious Bias in Machine Learning](https://www.infoq.com/presentations/unconscious-bias-machine-learning) by Rachel Thomas
- [Machinebias.org](http://machinebias.org/)
- [How To Stop Artificial Intelligence From Marginalizing Communities?](https://www.youtube.com/watch?v=PWCtoVt1CJM) by Timnit Gebru
- [Awesome Machine Learning Interpretability](https://github.com/jphall663/awesome-machine-learning-interpretability)
- [Weapons of Math Destruction](https://www.amazon.com/Weapons-Math-Destruction-Increases-Inequality/dp/0553418831/ref=as_li_ss_tl?ie=UTF8&qid=1532024714&sr=8-1&keywords=weapons+of+math+destruction&linkCode=sl1&tag=theopesoudats-20&linkId=6d7cfcea0ff64900854da08d3512a94e) by Cathy O'Neil
.center[]
---
class: inverse, center, middle
# Thanks!
</textarea>
<script src="https://remarkjs.com/downloads/remark-latest.min.js"></script>
<script src="libs/macros.js"></script>
<script>var slideshow = remark.create({
"highlightStyle": "github",
"highlightLines": true,
"countIncrementalSlides": false
});
if (window.HTMLWidgets) slideshow.on('afterShowSlide', function (slide) {
window.dispatchEvent(new Event('resize'));
});
(function() {
var d = document, s = d.createElement("style"), r = d.querySelector(".remark-slide-scaler");
if (!r) return;
s.type = "text/css"; s.innerHTML = "@page {size: " + r.style.width + " " + r.style.height +"; }";
d.head.appendChild(s);
})();</script>
<script>
(function() {
var i, text, code, codes = document.getElementsByTagName('code');
for (i = 0; i < codes.length;) {
code = codes[i];
if (code.parentNode.tagName !== 'PRE' && code.childElementCount === 0) {
text = code.textContent;
if (/^\\\((.|\s)+\\\)$/.test(text) || /^\\\[(.|\s)+\\\]$/.test(text) ||
/^\$\$(.|\s)+\$\$$/.test(text) ||
/^\\begin\{([^}]+)\}(.|\s)+\\end\{[^}]+\}$/.test(text)) {
code.outerHTML = code.innerHTML; // remove <code></code>
continue;
}
}
i++;
}
})();
</script>
<!-- dynamically load mathjax for compatibility with self-contained -->
<script>
(function () {
var script = document.createElement('script');
script.type = 'text/javascript';
script.src = 'https://mathjax.rstudio.com/latest/MathJax.js?config=TeX-MML-AM_CHTML';
if (location.protocol !== 'file:' && /^https?:/.test(script.src))
script.src = script.src.replace(/^https?:/, '');
document.getElementsByTagName('head')[0].appendChild(script);
})();
</script>
</body>
</html>