Ethical dilemmas in data science and analytics

About half a year ago, when my article on ethics in data science appeared on Harvard Business Review, the second half of the piece was dropped. In the omitted section, I had designed a poll inviting readers to opine on several ethical dilemmas facing real-world data analysts.

In April, Manoj Chari read my HBR article and kindly invited me to speak on the subject at the INFORMS Analytics Conference, and that provided the perfect occasion to dust off the poll. Robert Magliara and Jeff Cohen helped set up the PollEverywhere tool that is simple to use during the live session.

The attendees of the Conference are a mix of practitioners and academics, all involved with data analytics in some way. I presented a series of scenarios; the audience was polled about where they stood on each issue. About 22-25 people voted in each instance.

In each scenario bar one, the majority of respondents disagreed with what the hypothetical company or person was doing with the data.

The scenario that barely missed majority dissent concerned marketing services that link your email address to your browser cookies so that marketers can find you anywhere online. Some in the audience recognized that such services actually exist (in fact, all the scenarios are based on real-world examples). Arguing "someone is already doing it" is quite common in ethical debates but, to me, unsatisfactory. Another possible explanation is that the dislike for advertisers is not as strong as one might think.

The "I don't care" answers are intriguing. Roughly a quarter of respondents selected the neutral position for most scenarios. There were two scenarios in which opinion was polarized, with almost no one selecting "Don't Care".

One of the polarizing issues concerned trading of sensor data between auto-makers and auto-insurers, while the other described fake antennas used to compile databases of your locations and movements.

***

Here are the six scenarios that I offered up to the live audience:

Should BuzzCut position the banner ads to benefit from accidental clicks?

Yes, Just Do it! (18%)

No, Don’t Even Think About It (55%)

I Don’t Care (27%)

Should the automaker sell its data to the insurer?

Yes, Just Do It! (9%)

No, Don’t Even Think About It (87%)

I Don’t Care (4%)

Should a business run targeted ads using Middle’s database?

Yes, Just Do It! (29%)

No, Don’t Even Think About It (46%)

I Don’t Care (25%)

Should ChitChat uncover the names of users who use pseudonyms?

Yes, Just Do It! (14%)

No, Don’t Even Think About It (68%)

I Don’t Care (18%)

Should HearNoEvil be allowed to run such a business?

Yes, Just Do It! (20%)

No, Don’t Even Think About It (80%)

I Don’t Care (0%)

Should Bill help his godson out?

Yes, Just Do It! (14%)

No, Don’t Even Think About It (64%)

I Don’t Care (23%)

 

Comments

[Dan Wigglesworth]

There appears to be two purposes to the poll.

1. To prompt readers to reflect on ethical dilemmas in which they might find themselves (useful!) and

2. To get insight into the collective mind of a sample of CAPs -- perhaps with a view to discovering how one stands relative to peers. If I find that I appear to be an "outlier", I should consider rethinking my position (with caution).
   

Considering the latter point, it might be instructive to know how a given respondent varies from response to response. For example, is it the same 2, 3 or 4 respondents who chose "Yes, Just do it!" in almost every case? Or was there more variability? Alternatively, were there any respondents who were consistently in the "Don't even think about it?" category who then switched to the "Yes, Just do it?" category? If so, in which scenario's did this occur?

[maryjane]

great information.

[maryjane]

great information.

[Samuel Kazungu Muramba]

Very informative article for ethics committee members

[Victor]

Great post. Thanks for sharing

[Martin Buuri Kaburia]

Good study

[David]

Great Post!