AI Ethics Case Study [padlet.com/kayjanw/week4b] by KJ Wong https://padlet.com/kayjanw/week4b Activity Duration: 20 minutes en-us 2021-06-29 00:32:37 UTC 2022-11-25 07:30:22 UTC hello@padlet.com Instructions kayjanw https://padlet.com/kayjanw/week4b/wish/1629121451 1. Quick summary of the case
2. What are the discrimination and bias involved?
3. What caused the discrimination and bias?
4. How can the discrimination and bias be mitigated?

* Assign a dedicated note-taker because each post can only be edited by one person at a time
* Raise your hand in breakout room if you need any help

]]> 2021-06-29 00:32:37 UTC https://padlet.com/kayjanw/week4b/wish/1629121451 Article 1 kayjanw https://padlet.com/kayjanw/week4b/wish/1629121452 This Nature article reports on a research study published in Science that examines racial bias in AI algorithms:

Reference:
Dissecting Racial Bias in an Algorithm Used to Manage the Health of Populations, Science (366:6464), October 2019, Obermeyer et al. 
https://www-science-org.libproxy1.nus.edu.sg/doi/10.1126/science.aax2342
]]> 2021-06-29 00:32:37 UTC https://padlet.com/kayjanw/week4b/wish/1629121452 Article 2 kayjanw https://padlet.com/kayjanw/week4b/wish/1629121453 Machine Bias – AI Predicting Future Criminals is Biased Against Blacks

ProPublica is a nonprofit organization based in New York City that produces investigative journalism
This ProPublica article reports on potential bias in the COMPAS algorithm:
Correctional Offender Management Profiling for Alternative Sanctions
Used in US court systems to predict the likelihood that a defendant would become a recidivist


]]> 2021-06-29 00:32:37 UTC https://padlet.com/kayjanw/week4b/wish/1629121453 Article 3 kayjanw https://padlet.com/kayjanw/week4b/wish/1629121454 This The Verge article reports on a research study published in an ACM conference that examines bias in online advertising:

Reference:
Discrimination through Optimization: How Facebook's Ad Delivery Can Lead to Biased Outcomes, Proc. ACM HCI, November 2019, Ali et al.
https://dl.acm.org/doi/abs/10.1145/3359301
]]> 2021-06-29 00:32:37 UTC https://padlet.com/kayjanw/week4b/wish/1629121454 Group A kayjanw https://padlet.com/kayjanw/week4b/wish/1629121455 1.  An algorithm widely used in US hospitals to allocate health care to patients has been systematically discriminating against black people

2. Less money is spent on Black patients who have the same level of need, and the algorithm thus falsely concludes that Black patients are healthier than equally sick White patients.

3. The bias arises because the algorithm uses health care cost accured in a year to predict patients who are of higher risk category and requiring more personalize care. The bla osts rather than illness, but unequal access to care means that less money was spent caring for Black patients than for White patients.

4. 

]]> 2021-06-29 00:32:37 UTC https://padlet.com/kayjanw/week4b/wish/1629121455 Group B kayjanw https://padlet.com/kayjanw/week4b/wish/2231285790 1. An algorithm widely used among US hospitals and insurers to allocate health care to patients has been systematically discriminating against black people. Algorithm was less likely to refer black people than white people who were equally sick to programmes that aim to improve care for patients with complex medical needs.

2. Assign Black patients the same level of risk than White patients. Racial bias reduces the number of Black patients identified for extra care by more than half. Bias occurs because the algorithm uses health costs as a proxy for health needs.

3. Algorithm assigned risk scores to patients on the basis of total health-care costs accrued in one year without the data of the Patient's health condition

4. Instead of relying on AI independently, human should investigate the data source and develop technical ways such that the result will be fair.]]> 2022-06-27 08:55:29 UTC https://padlet.com/kayjanw/week4b/wish/2231285790 Group C kayjanw https://padlet.com/kayjanw/week4b/wish/2231286059 1.  Using AI to do risk assessments and predict future criminal acts and re-offences

2. race, nationality, skin colour 
3. questions skewed towards race, nationality, family background, job etc

4. feeding more relevant or more balanced data and co-relation]]> 2022-06-27 08:55:43 UTC https://padlet.com/kayjanw/week4b/wish/2231286059 Group F kayjanw https://padlet.com/kayjanw/week4b/wish/2231286198 1. 

2. 

3. 

4. ]]> 2022-06-27 08:55:56 UTC https://padlet.com/kayjanw/week4b/wish/2231286198 Group E kayjanw https://padlet.com/kayjanw/week4b/wish/2231286267 1. Facebook's ad delivery leads to biased outcomes due to the platform optimisation which defers from the advertiser target audience

2. What are the discrimination and bias involved? gender, racial, advertiser's budget and ads contents

3. What caused the discrimination and bias? platform ads delivery optimization

4. How can the discrimination and bias be mitigated? Facebook is likely reinforcing social inequalities, changes to ad targeting tools
Suspend targeted advertising on posts for jobs or housing.
Change its targeting system to actively counter bias
Shunt these listings to a separate system, like the housing ad database]]> 2022-06-27 08:56:02 UTC https://padlet.com/kayjanw/week4b/wish/2231286267 Group D kayjanw https://padlet.com/kayjanw/week4b/wish/2231286369 1. There are racial disparities in risk scores in black and white defendants. Some states are using a risk assessment tool to minimise the bias in the system.
2. race and skin color

3. 

4. ]]> 2022-06-27 08:56:11 UTC https://padlet.com/kayjanw/week4b/wish/2231286369