Ethics and Responsibility. Definitions Etiology –Greek "ethos" meaning "character" –Latin Moral "mos" meaning "custom" Ethics The rules that govern what.

Ethics and Responsibility

Definitions

• Etiology– Greek "ethos" meaning "character" – Latin Moral "mos" meaning "custom"

• EthicsThe rules that govern what is right and what is wrong for a person to do

• MoralsThe principles of right and wrong based on a standard (society or religion), on which ethics are based

Ethical Considerations

• Privacy• Data security• Accuracy• Research

Technology and Ethics

• Ethical considerations follow technological innovations

• Stick & stones• Nuclear fission• Information technology

– File sharing– Passwords– Data storage

Codes of Ethics

• ACM Code of Ethics• IEEE Code of Ethics• Medical Code of Ethics• Manataka American Indian Council

• Greek Fraternity Code of Ethics

Application of Ethics

• Ethics implies social considerations

• Involves– Privacy– Data security– Accuracy– Research

DI and Privacy• Large files may contain private information

• Data could be used long after its original collection date

• Data could be used beyond its collection purpose

• The purpose of knowledge discovery may not be known until some pattern is revealed in the data

• The information revealed by DI may be inappropriate

Protecting Privacy

• Privacy is contextual and a based on individual perceptions so a global assurance of privacy cannot be achieved

• Steps to enhance privacy protection– Anonymization of personal data – Provide to the user a method to review their personal data that is to be used in DI and decide for themselves

Privacy

• The type and amount of information a person is willing to share defines their privacy

• Notion of privacy violation• Organization of Economic Cooperation and Development guidelines – 1980

• Family Educational Rights and Privacy Act (FERPA - 1974)

• Health Insurance Portability and Accountability Act (HIPAA - 1996)

Problems with Privacy

• Current practices fall short of protecting privacy– People are unaware of the need to communicate with data holders

– People must be assertive and proactive

– The government as Big Brother•The Patriot Act

DI and Databases

• Two positions to consider:– DI operations authorized by an individual or organization that hold full access to the data

– DI operations unauthorized to mine the data, but have access for other reasons

Protecting Databases

• Authorization– Single level vs Multi-Level Security

• Encryption• Auditing• Precautions

– Mining only one security level can prevent inference from less sensitive data to more sensitive data

– Render the data useless for mining• Introduction of noise in the data• Introduction of instability in the data

DI and Data Accuracy

• DI uses data from many diverse, possibly external data sources

• Initial data quality cannot be known– Noisy, obsolete, inaccurate, incomplete

• Expired data can lead to inaccurate patterns discovered

Data Accuracy

• Inaccuracies are difficult to correct

• Expired data is undetectable until a person is affected by applying the patterns discovered in the data

• Adopt data quality management• Correct errors in data with expediency

• Frequent data cleaning

DI and Research

• Science is founded in truth and relies on ethical behavior in practice

• Clearly defined ethical guidelines for DI are not yet stated

• Consider ethical strategies for– Data collection, data storage, retention, authorship, publication, supervision of students and research assistants, disclosure, misconduct rules

– Human subjects

User-defined Sensitivity

Factors• Reference:– Wahlstrom K, Roddick J. On the Impact of Knowledge Discovery and Data Mining. 2nd Australian Institute of Computer Ethics Conference (AICE2000), Canberra, 2001.

– Users are the most qualified to qualify the sensitivity of their own data within context

– Provide users with a tool to specify sensitivity level when data is being collected

– Use sensitivity value with data in DI methods

DI Code of Ethics

• There isn’t one.• Should there be?