Chapter 3: Questions & Answers

Original Source Here

1. Does ethics provide a list of “right answers”?

Ethics: A branch of philosophy that involves systematizing, defending, and recommending concepts of right and wrong behavior. It seeks to resolve questions of human morality by defining concepts such as good and evil, right and wrong, virtue and vice, and justice and crime. It also has 3 major areas of study which are meta-ethics, normative ethics, and applied ethics.

  • Ethics doesn’t provide a list of “right answers” for solving problems. It provides a set of principles that can help mitigate confusion, clarify an issue, and identify some clear choices. It can also help identify several “right answers” but each person must come to their own conclusions.

2. How can working with people of different backgrounds help when considering ethical questions?

Research suggests the problem-solving skills of a diverse group outperform those of groups comprised of the most talented individuals. It adds different perspectives from people with different identities and experiences that may have privileged access to relevant insights and understandings. It also helps form policies and develops research that better caters to the people’s needs.

3. What was the role of IBM in Nazi Germany? Why did the company participate as it did? Why did the workers participate?

  • The role of IBM was to supply the Nazis with data tabulation products that tracked the extermination of Jews and other groups on a massive scale. It also provided training and maintenance onsite at the concentration camps.
  • The company’s President, Thomas Watson, was accused of participating with the Nazis for profit. He personally approved the request to supply the Nazis with specialized machines. He also visited the Third Reich multiple times, dined with Hitler, and accepted the “Service to the Reich” medal.
  • The workers may have participated because they were following orders. They were working-class men that were likely trying to live ordinary lives, care for their families, and do well at their jobs. They were also likely influenced by things like group dynamics of conformity, deference to authority, and the altering of moral norms to justify their actions.

4. What was the role of the first person jailed in the Volkswagen diesel scandal?

Volkswagen Diesel Scandal: A scandal where Volkswagen admitted to cheating on government emissions tests. It was revealed that they installed illegal software that only turned on the pollution controls when the vehicles were being tested to temporarily reduced the emissions by up to 40 times.

  • James Liang was an engineer at Volkswagen. He knowingly designed the software that detected when the vehicles were tested and changed the engine settings to temporarily improve performance. He was also sentenced to 40 months in prison and ordered to pay a $200,000 fine.

5. What was the problem with a database of suspected gang members maintained by California law enforcement officials?

In 2016, a state audit revealed that the CalGang database contained many errors that diminished its crime-fighting value. It found flaws in the system such as little oversight, lack of transparency, policy violations, and trouble justifying why some people were added to the system. It also had no process in place to correct mistakes or remove people after they have been added.

  • It identified 42 gang members who were less than one year old.
  • It identified 13% of people that were included inappropriately.
  • It identified 600+ people whose files should have been removed

6. Why did YouTube’s recommendation algorithm recommend videos of partially clothed children to pedophiles, even though no employee at Google had programmed this feature?

The YouTube recommendation system is designed to increase the amount of time that people spend watching videos. It automatically creates playlists that contain video recommendations for people based on what they watch and what similar people have watched. It also continuously optimizes that metric which produces very popular playlists without discrimination.

7. What are the problems with the centrality of metrics?

The problem with the centrality of metrics like YouTube’s decision to optimize their recommendation system to maximize the amount of time that people spend watching videos is it creates all kinds of extreme situations where people will search for, find, and exploit these situations and feedback loops.

8. Why did not include gender in its recommendation system for tech meetups?

Meetup didn’t include gender in its recommendation system because they felt it would be better to recommend meetups to people regardless of their gender. It noticed men expressed more interest than women in technology meetups. It also concluded that it would create a feedback loop that would cause even fewer women to find out about and attend these meetups.

9. What are the six types of bias in machine learning, according to Suresh and Guttag?

Historical Bias: A bias that occurs when the data that’s used to train the model no longer represents the current reality. It occurs even when the measurement, sampling, and feature selection are done perfectly because the people, processes, and society that are involved have existing biases.

Measurement Bias: A bias that occurs when the wrong features and labels are measured and used to train the model. It occurs when the model makes errors because the wrong thing is measured, the right thing is measured the wrong way, or the measurement is incorporated into the model incorrectly.

Aggregation Bias: A bias that occurs when the model can’t distinguish between the different groups in the heterogeneous population. It occurs because the model assumes the mapping from inputs to labels is consistent across all the groups when they are usually different for different groups.

Representation Bias: A bias that occurs when the model fails to generalize well because one or more parts of the population are under-represented. It occurs because the model identifies a clear relationship in the sample and assumes the relationship is an accurate representation of the population.

Evaluation Bias: A bias that occurs when the data that’s used to measure the performance of the model isn’t representative of the population or isn’t appropriate for the way the model will be used. It occurs because the data contains biases that inaccurately evaluate the performance of the model.

Deployment Bias: A bias that occurs when the model is incorrectly used to solve a problem that’s different than what it was designed to do. It occurs because the model is being used to perform a task individually that actually takes moderation by institutional structures and human decision-makers.

10. Give two examples of historical race bias in the US.

Correctional Offender Management Profiling for Alternative Sanctions (COMPAS): A tool that’s used by U.S. courts to assist judges with their sentencing and bail decisions. It predicts the likelihood a defendant will recommit a criminal offense. It also provides an individual risk assessment score for general offenses, violent offenses, and pretrial misconduct.

  • In 2016, an investigation revealed the COMPAS algorithm contained clear racial biases in practice. It found that black Americans were twice as likely as white Americans to be labeled as high risk when they didn’t re-offend. It also found that white Americans were much more likely than black Americans to be labeled as low risk when they did re-offend.
  • In 2012, a study revealed that all-white juries were 16% more likely to convict a black defendant than a white one. It found black defendants were convicted 81% of the time where white defendants were convicted 61% when there were no black jurors in the jury pool. It also found the conviction rate reduced to 71% for black defendants and raised to 73% for white defendants when at least one black juror was in the jury pool.

11. Where are most images in ImageNet from?

The ImageNet the dataset was assembled using images that were scraped from image search engines and Flicker. It contains images that primarily reflect the U.S. and Western countries because most of the images on the internet were uploaded from those countries at the time the dataset was assembled.

12. In the paper “Does Machine Learning Automate Moral Hazard and Error?” why is sinusitis found to be predictive of a stroke?

Sinusitis was mistakenly found to be a predictor of stroke because the model was trained using an inadequate type of training data. It was trained using biological and behavioral data that’s indirectly related to identifying stroke such as the documented stroke-like symptoms, whether the patient decided to seek medical care, and whether they were tested and diagnosed by a doctor. It also didn’t use data that are directly related to identifying stroke such as the biological signature of blood flow restriction to the brain cells.

13. What is representation bias?

Representation Bias: A bias that occurs when the model fails to generalize well because one or more parts of the population are under-represented. It occurs because the model identifies a clear relationship in the sample and assumes the relationship is an accurate representation of the population.

14. How are machines and people different, in terms of their use for making decisions?

Humans and machines aren’t used interchangeably for making decisions.

  • Machines are assumed to be objective and error-free.
  • Machines can be implemented without an appeals process in place.
  • Machines can be implemented at scale.
  • Machines are much cheaper than the cost of human decision-making.

15. Is disinformation the same as “fake news”?

Disinformation: False or misleading information that’s presented for the purpose of manipulation. It usually originates from the intention to cause economic damage, manipulate public opinion, or generate profits. It also contains exaggerations and seeds of truth that are taken out of context.

Fake News: False or misleading information that’s presented as legitimate news. It usually originates from the intention to damage the reputation of a person or entity or to generate profits through online advertising. It also contains purposefully crafted, sensational, emotionally charged, fabricated, and or misleading information that mimics the form of mainstream news.

16. Why is disinformation through autogenerated text a particularly significant issue?

Text generation can be used to spread disinformation rapidly. It can create compelling content on a massive scale with far greater efficiency and lower barriers to entry than ever before. It can also be used to perform socially harmful activities that rely on text such as spam, phishing, abuse of legal and government processes, fraudulent academic writing, and pretexting.

17. What are the five ethical lenses described by the Markkula Center?

Ethical Lenses: A tool that helps technology companies embed ethical considerations into their workflow to promote the development of ethical products and services. It contains theories that are widely used by academic and professional ethicists in the context of Western philosophical thought.

  • The Rights Approach: Which option best respects the rights of all who have a stake?
  • The Justice Approach: Which option treats people equally or proportionately?
  • The Utilitarian Approach: Which option will produce the best and do the least harm?
  • The Common Good Approach: Which option best serves the community as a whole, not just some members?
  • The Virtue Approach: Which option leads me to act as the sort of person I want to be?

18. Where is policy an appropriate tool for addressing data ethics issues?

Data ethics policies become a priority to companies when there are heavy financial and legal consequences that are imposed through regulations and laws. It can become necessary to protect the public through coordinated regulatory actions when data ethics issues are violating human rights and are impossible to solve through individual actions like purchase decisions.


Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot

%d bloggers like this: