2024-01-01-154304 Ethics and data science - a hippocratic oath and other things

[2024-01-01 Mon 15:43]

Related: DataScienceJournal CodeJournal DataScienceEthics? PhilosophyJournal DataGovernance? <<<CloudComputing? MyCommentary

WIRED has an article out that highlights some interesting things about discussions related to Ethics in AI and data science. The meeting hashed out a bunch of principles which are meant to be starting points and guidelines for policy makers to think about data ethics. One of the main questions asked was whether practitioners should follow some kind of hippocratic oath, considering that the practitioners are best positioned to implement sound data ethics in the whole process.

There was atleast one person who thought the principles stated were so general that nobody would disagree with them. The concern was also there that companies would be happy to agree to these principles and then continue business as usual. There was an interesting example where data scientists in AT&T refused to work on a project since they felt the vendor involved was not ethical - and while that was accepted, the project just moved on with different data scientists who did not have that problem. In such cases, it feels like the so called hippocratic oath is 'optional' which detracts significantly from the value and impact of the oath.

DJ Patil himself expressed his thought that society cannot rely on tech companies to think carefully about the effects of their data fuelled products. It was mentioned that historical evidence suggests that tech companies become contrite or take measure against these issues only when under sufficient pressure, and this means either financial, or from the Government. i.e. measures like the GDPR which are government driven would be more effective than an oath taken by an individual practitioner.

What good would an oath do, if only a handful of employees in a large company or team decide to follow it? Is this an area where one would be comfortable trusting the honor system? I for one would not want to trust an honor system.

I wonder though, irrespective of following an oath or things being enforced by a formal GDPR esque policy - is there really a way as of today to enforce any policy with respect to data? Who exactly is checking that the rules have been followed?

All that being said, one has to start 'somewhere'. A hippocratic oath is better than having no oath, and could be a signal to the entire 'culture of data science', imbibed from the fundamental educational stage. Some guidelines existing to help policy makers is better than no guidelines. Enforcing may be a thing that could be looked at, after figuring out what policies need mandatory adoption.