Responsible AI Challenge

	A	B	C	D	E
1	Fairness \| Privacy \| Security \| Explainable
2	What makes AI, responsible?	What motivated this thought?	Resources	Any interesting points from it

3	Is dataset representative of every possibility	Datasheets for datasets help us ask questions on bias in data	https://www.youtube.com/watch?v=UEECKh6PLhI&list=PLQY2H8rRoyvzuJw20FG82Lgm2SZjTdIXU&index=20	think about fairness in every stage of the pipeline
4	Bias aware loss function	Make the loss function penalize unfairness	https://www.coursera.org/learn/nlp-sequence-models/lecture/zHASj/debiasing-word-embeddings	Diminishing Gender / Ethenicity / Orientation bias in word embeddings	aif360 algorithms - Prejudice remover - Adversarial Debiasing
5	Translate features so that are less biased	For example : Age -> experierience wrt a skill	https://www.infoq.com/presentations/unconscious-bias-machine-learning/	Analyzing & Preventing Unconscious Bias in Machine Learning Solutions from 24 min Questions to ask for fairness from 31m 20s
6	Can we have Non representative test data	We always look for how well the model is duing with test data . If the test data follows similar distribution as the train data . The bias in the model can be translated into test results as well	https://responsible-ai.devpost.com/	Links for tools measuring fairness
7	Error rates on different sub-groups	Ensures that dataset is diverse or not	https://microsoft-my.sharepoint.com/:p:/g/personal/nipasuma_linkedin_biz/EZGTpZq1TsFGjv3O43XURJEB3Qh1MbnLXwPEXdQ8UQZaMQ?e=F0pY9o	A good summary of explainable ML approaches
8
9
10
11
12
13
14
15
16
17
18
19
20
21			https://bluejeans.com/s/S7_4h/	Our kickoff BJ meeting recording
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