+ Create new tricision Login 

Login

or


Forgot password?

First visit? Sign-up now!

By continuing you accept the Terms & Conditions and Privacy Policy.

time is up

ML and the Law: Panel 1 - nearer term issues

Please add and/or vote on topics you feel interesting

(please ignore the pros and cons field)


saved
Ideas
Pros and cons
 
Votes
Shouldn't the law require the population to understand ML basics, rather than require giving simplified (= often incorrect) explanations - cf. legal obligation of parents, e.g. in Europe, to scholarize their children (to be able to read among other goals)
 
Is this a joke?
5
 
no, given the technology proliferation trend, it's not very far-fetched having ML basics in schoo... more
1
 
7
Vote

Hinda, K, Y and 4 more

Why, specifically, is machine learning of particular concern --- cf. general (complex) IT systems/software?
5
Vote

CM, Hinda, T and 2 more

Machine learning applications in hiring, credit, sentencing and beyond are likely to be biased and discriminatory, but humans decision makers are often biased and discriminatory as well. In tort law, it can be difficult to prove that a human was acting in a discriminatory way. But with ML-based d... more
by Robert McGibbon
4
Vote

CM, Cody, F and 1 more

The EU's GDPR requires that subjects have the right for a "meaningful information about the logic involved". Assume that the prediction model was learned from obfuscated (e.g. encrypted) data, where the learner is different from the data handler (the data handler does not know the learning algori... more
4
Vote

CM, Lisa Koch, Hinda and 1 more

How does the law define what a 'sufficient expanation' is? What if the concened individual still 'doesn't get it'?
4
Vote

D, Hinda, William and 1 more

The EU's GDPR, as mentioned in the first round of talks, includes potentially large fines for failure to comply with its algorithmic transparency and accountability requirements within a few years. Given that academic research on this subject (e.g. the interpretability of neural nets) seems more ... more
3
Vote
AR
K
M'
What does - or what should - fair mean?
3
Vote
A
A
Cody
are there efforts to suggest clearer/better laws taking into account new technology which has privacy /fairness implications? how can we contribute to such efforts?
by Jan Ramon
2
Vote
Cody
Jan Ramon
Is tort law that assigns liability to developers a good approach to regulating AI?
2
Vote
Dylan
ian
What applications or uses of ML would serve as good legal "test-cases"?
2
Vote
Hinda
Robert
Given the engineering cost of collecting data and the cost of obtaining consent to collect additional data, why wouldn't a practitioner err on the side of collecting as much as is possible and conceivably needed in the future?
 
I would think that the question of collection should be 'what, in expectation, is the marginal va... more
by Matt Gershoff
 
Perhaps, but if the initial marginal cost is near zero (for adding an additional attribute), but ... more
 
Reasonableness is the issue, what is not reasonable, IMHO, is to expect that all uncollected data... more
by Matt Gershoff
1
Vote
K
What is meant by a significant decision for the regulation around automated decisions under GDPR. Any examples of a decision that wouldn't be covered as well as one that would be?
1
Vote
M
What role should uncertainty play?
1
Vote
A
How to properly separate and define the legal responsibility when deploying an AI system?
1
Vote
R
Independent of current laws, what arguments do you think there are for imposing anti-discrimination controls on algorithmic decision making? If you think it valuable, what specific kinds or mechanisms of statistical discrimination do you think its most necessary to correct?
0
Vote
Are we holding machine learning to a higher standard than humans when we require that a decision made by a machine learning algorithm be explained?? That is, are humans always forced to explain their decisions (and moreover, in a coherent way)?
by Nishant Mehta
0
Vote
Given the engineering cost of collecting data and the cost of obtaining consent to collect additional data, why wouldn't a practitioner err on the side of collecting as much as is possible and conceivably needed in the future?
0
Vote

Comments

https://www.tricider.com/brainstorming/34pe1ibbTlJ