Studied by 27 people
Looks like no one added any tags here yet for you.
SEEP acronym
Security Economics ethics and Privacy
5 V's of Big Data
Volume: All the numbers, Walmart handles over 1 million transactions per hour
Velocity: How fast is data being added, Clickstreams an ads, machine to machine processes
Variety:What’s in the data, not just numbers, strings,dates but looking at different types. Less predictable, less structured, like google maps has directions, restaurant reviews.
Veracity: Is the data any good? The quality, authenticity, and validity of the data
Value: What’s the point of the data? What an organization gains for retaining or acquiring the data
FIPS Federal Information Processing Standards
Collection Limitation: The collection of personal information should be limited, should be obtained by lawful and fair means, and with individual consent
Data Quality Purpose Specification: Quality of data should be kept to specific standards
Security Safeguards: Personal data should be protected by reasonable security safeguards against such risks as loss or unauthorized access, destruction, use, modification, or disclosure of data
Use Limitation: Personal data should not be disclosed, or made available or other used for purposes other than those specified
Purpose Specification: The purpose for which personal data are collected should be specified not later than at the time of data collection
Openness: There should be a general policy of openness about developments, practices and policies with respect to personal data
Accountability: A data controller should be accountable for complying with measures which give effect to the principles
Individual Participation: An individual has the right to obtain from a data controller, to have communicated to him data relating to him, to be given reasons if a request made is made denied to be able to challenge such denial, to be challenged data relation to him and if the challenge is successful to have the data erased, rectified, completed, or amended.
Utiliarianism
Focused on outcome; actions right or wrong depend on consequences caused
Deontology
Actions right or wrong depends on its conformity to moral norms and rules
Virtue Ethics
Goal of life is happiness and the means to attain it is through a virtuous character
What is a bias?
Preference towards toward a particular viewpoint
Goals of algorithms
Process or set of rules followed by calculations done by computer
Psuedocode
Cross between human language and a programming language
Technology determinists
Technology itself is a primary factor and is at fault
Social determinists
Society is the primary actor and is at fault; blame use of program and design
Preexisting bias Bias
existing prior to creation of system
Summarize the Stanford Vaccine Algorithm Case
Algorithm was designed to give COVID vaccine that prioritized younger and older people first rather then front line workers.
Summarize the Medical Algorithm Favoring White Patients Case
Algorithm was race blind and ranked patients on how much they would cost the health care system. It scored white patients as equally at risk of future health problems as Black patients despite them suffering more serious diseases. It was based off incurred medical costs
Disparate Treatment
differentially treating a protected class based on status
Disparate Impact
Discriminatory effects even when applying "neutral" rules
Why is fairness in algorithms difficult?
If the data being used is s biased from the start the machine has no way of recognizing the problem
Summarize the Bias in Criminal Risk Scores Case
Analysis of bias against black defendants in criminal risk scores showed that the disparity can be addressed if the algorithm focuses on the fairness of outcomes. Black defendants were more than twice as likely to be marked higher risk by the COMPAS formula.
Define employee monitoring
Monitor to protect company secrets and data, can conduct searches as long as employer has notified of privacy expectations
Summarize the Bias in Credit Scores Case
Predictive tools used for mortgage lending are less accurate for minorities. Due to them having less data in their credit histories, the system was more likely to decline the loan request.
Define employee monitoring
Monitor to protect company secrets and data, can conduct searches as long as employer has notified of privacy expectations
Define workplace communication policies
Employers can monitor employee communications sent or received on company devices as long employer communicated dispelling the right of privacy in the system
What is Power in Surveillance?
the persistent tracking of an individual causes one to lose power against the actors that are surveilling
27. Panopticon
Social mechanism of control where people know that while they are not watched all the time, they may be watched at any time
Data sensitivity in healthcare
Contains sensitive data including positivity of disease, etc; what if employers or insurance companies gain hold of data?
Healthcare wearables
Health monitoring has positive implications for healthcare, insurance, and employers, but can lead to data leaks of private information
Room for technology and data to assist in healthcare
Can alert for potential health issues and disease diagnosis
Artifical Intelligence (AI)
Looks to mimic human intelligence
Machine Learning(ML)
Aims to teach machine to perform a specific tasks and provide results via analysis; can think for itself; subset of AI
General purpose ai
Difficult; brain has special and general functions not yet understood
Special purpose ai
Easier; chess playing programs, speech/image recognition
Turing test
a test for intelligence in a computer, requiring that a human being should be unable to distinguish the machine from another human being by using the replies to questions
Where do humans win against ai
Solutions to complex environments, adaptation and creativity, analogical reasoning, explaining reasoning, humor and social reasoning
Game theory vs. machine learning
Game theory studies strategic interaction between rational people and their outcomes, machine learning focuses on creating algorithms to learn patterns and make predictions from data
Examples of Game Theory
- Prisoner's Dilemma. - Based around two prisoners, who have the choice to either confess or deny a crime. - The consequences of the choice depend on what the other prisoner chooses.
Trolley Problem
Person A can take an action which would benefit many people, but in doing so, person B would be unfairly harmed. Under what circumstances would it be morally just for Person A to violate Person B's rights in order to benefit the group?
Need for explainable ai,
Black-box statistical predictions are inadequate, and explanations must be understandable to a non-specialist
Crowdsourcing as means of accountability
Internet workers do microtasks to solve computer-hard problems (Ex: Categorizing products, Wiki, content moderation)
Summarize the Houston School District and the Eval System Case
Houston ISD's use of a secret algorithm to evaluate teacher performance denied employees the ability to challenge their terminations. The algorithm was used to determine what teachers were evaluated, fired, and given bonuses. It did not explain how teachers could improve their scores, and did not allow for reevaluation.
Summarize the Cheating Detection Case,
Algorithmic use of overanalyzing normal eye/body movements, finishing the exam too fast, noise, etc. Students raised concerns about the unfairness of adding extra stress on students to catch a small number of cheaters.
Summarize the Humans Getting the Blame Case
When algorithms are causing failures, the people behind them are getting the blame. Even in a highly automated system where humans have limited control of the behavior, they still face the consequences of its failure.
Gamification v actual games
Gamification is incorporating game elements to engage and motivate ppl (badges, leaderboards) to enhance and improve user engagements; Games have defined rules, goals and are used for entertainment, enjoyment and competition
Examples of gamification
Fitness apps with challenges and awards, Duolingo language learning with streaks and points
Main components to gamification
Goals and objectives, rewards and incentives, competition and collaboration, progression
Ethical concerns of gamification
Addiction and dependency, manipulation and exploitation, privacy issues
Data behind video games
Player demographics, player behavior and engagement, monetization, VR, user reviews and feedback
Addiction in gamification
Reward systems, in-app purchases, time-based incentives, social incentives, overstimulation
Summarize the Uber Using Psychological Tricks Case
Uber utilized gamification techniques to influence drivers to work longer and harder. They used goal setting, automatically loading the next ride, encouragement, pop-ups, and income targeting.
Pillars of democracy
All people should have equal and effective chances to make their views known to others and all should have the equal ability and chance to vote with all votes counted equal
Public sphere
a place where society discuses the issues that affect everyone
Democracy and technology
Technology has the ability to allow conversation, discussion, and interaction due to unmediated access, global reach, and low barriers
Advantages over mass media
Offers the public sphere to reach common judgement, and is a predominant source of political information for the democratic society
Where is mass media lacking?
Limited tech literacy, cost of entry (hardware), limited education/literacy, language and accessibility barriers
Summarize the What Facebook Did to American Democracy Case
Since user engagement was the ultimate goal for FB, the platform did not remove or censor hate speech and insurrection-based postings. The algorithm based removal on whether the action would increase or decrease engagement, allowing the posts to been seen on more feeds. As content became near prohibited (lawful but awful), the more FB found users engaged with the content. The more awful the content, the more FB recommended the content
Velocity
Speed the data is added and integrated
Veracity
Quality of the data gathered; authenticity
Small data
Impact decisions in the present; ready for analysis
Big data
Large amounts of unstructured data; purpose is for data mining
Technical bias
Technical constraints or considerations
Emergent bias
Arises after some time after design completion
John Rawls' Theory of Justice
Self-interested individuals should enter into a social contract that minimizes harm to the weakest parties; basis of society is a set of tacit agreements
Robert Nozick Theory of Justice
Inequalities are a fact of life; questions interventions to change status quo and fix inequality
Michael Walzer Theory of Justice
Different spheres of distribution; justice should be understood under specific social spheres with its own principles of distribution
Data from one sphere is being used in an allocation decision that is measuring success or failure in another sphere
Expert system
Good for explanation, but poor with accuracy
Neural network
Good for accuracy but poor with explanations
Note 
Flashcard (23)