💽
Responsible Data Science
Created @May 6, 2024 4:37 PM
Class RDS
Created by Tieeny Chao
HC 1: Organization + Introduction
Data science —> accuracy and efficiency: what can we do with data
Responsible data science —> responsility: what we should and shouldn't do
with data
8 principles of AI ethics —> a set of values, principles, and techniques that
employ widely accepted standards of right and wrong to guide moral conduct
in the development and use of AI technologies.
1. Privacy —> AI systems should respect individuals' privacy, both in the use
of data and by providing impacted people with agency over their data and
decision made with it.
a. information collection: violated surveillance (observational) and
interrogation (directly asking)
b. information processing: what is done with the data once gathered.
i. should only be used for the right goals and consent obtained for.
c. information dissemination: data leaks and human gaze
d. invasion: disturb a private moment, home, thought and experience.
i. intrusion —> pop-up
ii. decisiononal interference —> convince someone to buy more food
than they want
Responsible Data Science 1
, 2. Accountability —> AI should include mechanism to ensure that
accountability for the impact of AI systems is appropriately distributed, and
remedies are provided.
3. Safety and security —> AI systems should be safe, performing as intended,
and also secure.
4. Transparency and explainability —> AI systems must be designed and
implemented to allow for oversight. Non-transparent or unexaplainable
outcomes can be caused by insufficient transparency.
5. Fairness and non-discrimination —> AI systems must be designed and
used to maximize fairness and promote inclusivity (against AI bias).
a. Statisitcal bias: a model is biased if it does not summarize the data
correctly.
b. Societal bias: if a dataset or model does not represent the world
“correctly”.
i. meaning: The words as it is or as it should be
6. Human control of technology —> Important decisions should remain
subject to human review
7. Professional responsibility —> ensuring that the appropriate stakeholders
are consulted and long-term effects are planned for by individuals that
develop and deploy AI systems.
8. Promotion of human values —> AI's ends and means by which it is
implemented, should correspond with our core values and promote
humanity's well-being.
_______________________________________________________
HC 2: The algorithm dimension
White-box vs. black-box algorithms
White-box model represenation —> known to the humans and
understandable how the model came into place.
linear regression, logistic regression, discriminant analysis
Responsible Data Science 2
Responsible Data Science
Created @May 6, 2024 4:37 PM
Class RDS
Created by Tieeny Chao
HC 1: Organization + Introduction
Data science —> accuracy and efficiency: what can we do with data
Responsible data science —> responsility: what we should and shouldn't do
with data
8 principles of AI ethics —> a set of values, principles, and techniques that
employ widely accepted standards of right and wrong to guide moral conduct
in the development and use of AI technologies.
1. Privacy —> AI systems should respect individuals' privacy, both in the use
of data and by providing impacted people with agency over their data and
decision made with it.
a. information collection: violated surveillance (observational) and
interrogation (directly asking)
b. information processing: what is done with the data once gathered.
i. should only be used for the right goals and consent obtained for.
c. information dissemination: data leaks and human gaze
d. invasion: disturb a private moment, home, thought and experience.
i. intrusion —> pop-up
ii. decisiononal interference —> convince someone to buy more food
than they want
Responsible Data Science 1
, 2. Accountability —> AI should include mechanism to ensure that
accountability for the impact of AI systems is appropriately distributed, and
remedies are provided.
3. Safety and security —> AI systems should be safe, performing as intended,
and also secure.
4. Transparency and explainability —> AI systems must be designed and
implemented to allow for oversight. Non-transparent or unexaplainable
outcomes can be caused by insufficient transparency.
5. Fairness and non-discrimination —> AI systems must be designed and
used to maximize fairness and promote inclusivity (against AI bias).
a. Statisitcal bias: a model is biased if it does not summarize the data
correctly.
b. Societal bias: if a dataset or model does not represent the world
“correctly”.
i. meaning: The words as it is or as it should be
6. Human control of technology —> Important decisions should remain
subject to human review
7. Professional responsibility —> ensuring that the appropriate stakeholders
are consulted and long-term effects are planned for by individuals that
develop and deploy AI systems.
8. Promotion of human values —> AI's ends and means by which it is
implemented, should correspond with our core values and promote
humanity's well-being.
_______________________________________________________
HC 2: The algorithm dimension
White-box vs. black-box algorithms
White-box model represenation —> known to the humans and
understandable how the model came into place.
linear regression, logistic regression, discriminant analysis
Responsible Data Science 2
