Data Science & Ethics
Les 1: Introduction
Responsible AI
If it was the best of times: If it was the worst of times:
§ Reduce risk § Data leaks
§ Reduce crime § Discrimination
§ Increase profitability § Digital pawns
§ Improve medical diagnosis § Filter bubble
§ Increased “good” § Increased “bad”
Data Science & Ethics
§ Ethics: “moral principles that control or influence a person’s behavior”
§ Moral: “connected with principles of right and wrong behavior”
Law: what you can do
Ethics: what you should do (should you be doing this even though it is lawfull?)
§ Responsible AI (formal definition): The design and application of AI that is
aligned with the moral values of society
Why Care?
1. Expected from society
§ Generation Z
• Born 1995 – 2010
• 90 million in the US
• Cares about social justices and ethics
Gen z is naturally starting to care more
2. Huge potential risks
§ Be aware of the risks and countermeasures
§ Risks for humans
1
, • Physical and mental well-being (self-driving)
• Privacy
• Discrimination
For example, AI determining whether
Risks for businesses
you receive a loan or not so there is a
• Reputational
potential risk, same with cv
• Financial (fines, less customers, …)
screenings using AI models
3. Potential benefits
§ Understanding ethical concerns and applying techniques to deal with
this, can improve the data, model and be a marketing instrument
• Remove bias in data: improve the accuracy and fairness of the model
• Explain predictions: improve the trust the model
• Ensure proper data gathering better data quality
• Part of a company’s brand (cf. 1. Expected from society)
4. The AI Act
Risk-based approach to regulate AI
Requirements for high-risk uses of AI
• Establish safeguards against biases in datasets
• Prescribed data governance and management practices
• Ability to verify and trace back outputs through the system’s life
cycle
• Including provisions for acceptable levels of transparency and
understandability for users of the systems,
• Appropriate human oversight over the system generally.
Summary:
§ Being ethical should be a Life goal in itself (philosophical goal)
§ Societal and business reasons:
1. Expected from society 3. Data science ethics can
2. Huge potential risks bring value
4. AI Act
2
,Goal of the course
Provide guidance and insight on deciding what is right and wrong when conducting
data science.
Machine Learning
§ Prerequisite
§ Machine Learning: automatic
Machine learning
extraction of knowledge from
algorithm Pattern or
data
§ Setting the scene with credit
scoring example
Definitions
§ Data Science Ethics: the domain of what is right and wrong when doing data
science
§ Responsible AI: the development and application of AI that is aligned with
moral values in society
Ethics Theories: Utilitarianism vs Deontological Ethics
§ Utilitarianism:
• =consequentialism, what is produced in the consequence of the act
• Action is moral if the consequence is moral, means to an end
• Justifies immoral things
§ Deontology:
• Not doing immoral actions
• Outcome does not matter
Aristole’s Nicomachean Ethic
§ Moral behavior can be found at the mean between two extremes: excess and
deficiency.
3
, ‘Golden mean’(between the two extremes) of data science ethics.
§ Between these 2 extremes:
• Deficiency: Not using any data at all
• Excess: Using all available data for any application, without any concern for
Therethe ethical
should concerns
always be a human who deals with these extremes!
Data Science Ethics Equilibrium
§ = A state of data science practices determined by the ethical concerns and utility
of data science.
1. Eg churn prediction
2. Eg CV sorting
High risks means doing more about
ethical concerns
Trolley Problem
Well-known thought experiment in ethics: Variants
utilitarian vs deontological ethics • What if single person is your child
or partner?
• What if you’re on a bridge, and
can only stop the trolley by pushing
a man standing next to you off the
bridge?
Example with self-driving car: should it kill the passengers or the person
walking on the road. Here AI ethics need to be discussed in great detail
4
Les 1: Introduction
Responsible AI
If it was the best of times: If it was the worst of times:
§ Reduce risk § Data leaks
§ Reduce crime § Discrimination
§ Increase profitability § Digital pawns
§ Improve medical diagnosis § Filter bubble
§ Increased “good” § Increased “bad”
Data Science & Ethics
§ Ethics: “moral principles that control or influence a person’s behavior”
§ Moral: “connected with principles of right and wrong behavior”
Law: what you can do
Ethics: what you should do (should you be doing this even though it is lawfull?)
§ Responsible AI (formal definition): The design and application of AI that is
aligned with the moral values of society
Why Care?
1. Expected from society
§ Generation Z
• Born 1995 – 2010
• 90 million in the US
• Cares about social justices and ethics
Gen z is naturally starting to care more
2. Huge potential risks
§ Be aware of the risks and countermeasures
§ Risks for humans
1
, • Physical and mental well-being (self-driving)
• Privacy
• Discrimination
For example, AI determining whether
Risks for businesses
you receive a loan or not so there is a
• Reputational
potential risk, same with cv
• Financial (fines, less customers, …)
screenings using AI models
3. Potential benefits
§ Understanding ethical concerns and applying techniques to deal with
this, can improve the data, model and be a marketing instrument
• Remove bias in data: improve the accuracy and fairness of the model
• Explain predictions: improve the trust the model
• Ensure proper data gathering better data quality
• Part of a company’s brand (cf. 1. Expected from society)
4. The AI Act
Risk-based approach to regulate AI
Requirements for high-risk uses of AI
• Establish safeguards against biases in datasets
• Prescribed data governance and management practices
• Ability to verify and trace back outputs through the system’s life
cycle
• Including provisions for acceptable levels of transparency and
understandability for users of the systems,
• Appropriate human oversight over the system generally.
Summary:
§ Being ethical should be a Life goal in itself (philosophical goal)
§ Societal and business reasons:
1. Expected from society 3. Data science ethics can
2. Huge potential risks bring value
4. AI Act
2
,Goal of the course
Provide guidance and insight on deciding what is right and wrong when conducting
data science.
Machine Learning
§ Prerequisite
§ Machine Learning: automatic
Machine learning
extraction of knowledge from
algorithm Pattern or
data
§ Setting the scene with credit
scoring example
Definitions
§ Data Science Ethics: the domain of what is right and wrong when doing data
science
§ Responsible AI: the development and application of AI that is aligned with
moral values in society
Ethics Theories: Utilitarianism vs Deontological Ethics
§ Utilitarianism:
• =consequentialism, what is produced in the consequence of the act
• Action is moral if the consequence is moral, means to an end
• Justifies immoral things
§ Deontology:
• Not doing immoral actions
• Outcome does not matter
Aristole’s Nicomachean Ethic
§ Moral behavior can be found at the mean between two extremes: excess and
deficiency.
3
, ‘Golden mean’(between the two extremes) of data science ethics.
§ Between these 2 extremes:
• Deficiency: Not using any data at all
• Excess: Using all available data for any application, without any concern for
Therethe ethical
should concerns
always be a human who deals with these extremes!
Data Science Ethics Equilibrium
§ = A state of data science practices determined by the ethical concerns and utility
of data science.
1. Eg churn prediction
2. Eg CV sorting
High risks means doing more about
ethical concerns
Trolley Problem
Well-known thought experiment in ethics: Variants
utilitarian vs deontological ethics • What if single person is your child
or partner?
• What if you’re on a bridge, and
can only stop the trolley by pushing
a man standing next to you off the
bridge?
Example with self-driving car: should it kill the passengers or the person
walking on the road. Here AI ethics need to be discussed in great detail
4
