1
00:00:00,932 --> 00:00:06,188
MICHELLE: In 2012, Facebook conducted a week
long research study on nearly 700,000

2
00:00:06,188 --> 00:00:11,280
unwitting users who had not consented
to a mental health experiment.

3
00:00:11,280 --> 00:00:14,940
They filled one group's news
feeds with positive messages, and

4
00:00:14,940 --> 00:00:17,390
another with negative content.

5
00:00:17,390 --> 00:00:19,870
At the end of the study,
Facebook concluded

6
00:00:19,870 --> 00:00:23,640
that the content had indeed influenced
people to be more positive or

7
00:00:23,640 --> 00:00:27,470
negative depending on
their news feed content.

8
00:00:27,470 --> 00:00:29,310
In an Atlantic article,

9
00:00:29,310 --> 00:00:34,020
experts argued the study didn't
violate Facebook's terms of service.

10
00:00:34,020 --> 00:00:38,800
But it certainly didn't comply with
the American Psychological Association's

11
00:00:38,800 --> 00:00:43,460
ethical principles and code of conduct
regarding consent and deceptive research.

12
00:00:44,690 --> 00:00:49,110
Author and scholar Shoshana Zuboff
cautions, "We can affect real world

13
00:00:49,110 --> 00:00:53,400
behavior and emotions without
triggering user awareness."

14
00:00:54,530 --> 00:00:59,420
Chamath Palihapitiya, a former VP of
growth at Facebook who scaled the social

15
00:00:59,420 --> 00:01:05,140
network to millions of users, resigned
from his position over ethical concerns.

16
00:01:05,140 --> 00:01:09,670
He said the short-term dopamine-driven
feedback loops that we have created

17
00:01:09,670 --> 00:01:13,750
are destroying how society works,
no civil discourse,

18
00:01:13,750 --> 00:01:18,600
no cooperation, misinformation, mistruth.

19
00:01:18,600 --> 00:01:23,600
In the documentary, The Social Dilemma,
he questioned the ethics of testing

20
00:01:23,600 --> 00:01:27,096
small feature changes,
calling them manipulative.

21
00:01:27,096 --> 00:01:31,170
A/B testing is commonplace as
software development these days.

22
00:01:31,170 --> 00:01:34,820
A certain percentage of users
are shown a tweaked interface

23
00:01:34,820 --> 00:01:37,180
to test the effectiveness of a change,

24
00:01:37,180 --> 00:01:40,180
while the remaining users
are shown the usual interface.

25
00:01:41,310 --> 00:01:46,250
When evaluating the ethics of user
research, where should the line be drawn?

26
00:01:46,250 --> 00:01:49,490
When should users be
informed of the research?

27
00:01:49,490 --> 00:01:52,040
It's difficult to determine hard rules.

28
00:01:52,040 --> 00:01:57,030
Manipulating emotions and behavior for
the worst is clearly unethical.

29
00:01:57,030 --> 00:02:02,610
Otherwise, I value A/B testing as a tool
for improving the user experience.

30
00:02:02,610 --> 00:02:07,290
When conducting other forms of user
research, such as user interviews,

31
00:02:07,290 --> 00:02:11,070
refer to IDEO's Little Book
of Design Research Ethics.

32
00:02:11,070 --> 00:02:13,390
There's a link in the teacher's notes.

33
00:02:13,390 --> 00:02:17,382
A few gems include:
"Clarify expectations upfront,

34
00:02:17,382 --> 00:02:20,360
get consent, and
protect participants' data."