I-202 Final Exam

Algorithmic Decision making

A computerized analysis that makes use of a mathematical model (a representation of a process) to analyze data and suggest an outcome

Ethics

A framework of right and wrong that tells us what we ought to do. A framework for action.

Process Perspective

Big data research requires new kinds of technology and analysis

Are data social

Data collection is social (context, what is in the data set, what is outside the data set, how data are cleaned, how data are made commensurable).

Are models social

Models reflect our understanding of how the world operates (or should operate). They reflect the state of knowledge and its limitations, the kinds of expertise incorporated, the values and biases of the model makers, and the limitations of the technologies used to create and run the model. Models generate outputs and the weight given to these outputs in human decision making is also a social process

Social-movement perspective

Vision of how Big Data and its related technologies can be used to transform society

Characteristics of Big Data

Volume, Velocity ,Variety

"Critical questions for big data" -Boyd and Crawford

We define Big Data as a cultural, technological, and scholarly phenomenon that rests on the interplay of: (1) Technology: maximizing computation power and algorithmic accuracy to gather, analyze, link, and compare large data sets. (2) Analysis: drawing on large data sets to identify patterns in order to make economic, social, technical, and legal claims. (3) Mythology: the widespread belief that large data sets offer a higher form of intelligence and knowledge that can generate insights that were previously impossible, with the aura of truth, objectivity, and accuracy Social systems are regulated by four forces: market, law, social norms, and architecture - or, in the case of technology, code. When it comes to Big Data, these four forces are frequently at odds. The market sees Big Data as pure opportunity: marketers use it to target advertising, insurance providers use it to optimize their offerings, and Wall Street bankers use it to read the market -Big data changes the definition of knowledge -Claims of objectivity and accuracy are misleading -Bigger data are not always better data -Taken out of context, big data loses its meaning -Just because it is accessible does not make it ethical -Limited access to big data creates new digital divide

Interplay of Big Data

(1) Technology:maximizing computation power and algorithmic accuracy to gather, analyze, link, and compare large data sets. (2) Analysis:drawing on large data sets to identify patterns in order to make economic, social, technical, and legal claims. (3) Mythology:the widespread belief that large data sets offer a higher form of intelligence and knowledge that can generate insights that were previously impossible, with the aura of truth, objectivity, and accuracy.

"AI now institute 2018 Symposium" Video (1-17)

-Google had been making AI systems for department of defenses drone surveillance program -ICE tampered with risk assessment algorithm to produce one result, to detain 100% of immigrants in custody -IBM built ethnicity detection feature for NYPD -Autonomous vehicles killed passengers and pedestrians -ACLU showed how amazon's facial recognition service incorrectly identified 28 congress members as criminals - 5 Themes emerging from accountability Material realities Labor Legal Strategies Inequality New coalitions

To Predict and Serve (Lum and Isaac)

-In late 2013, Robert McDaniel - a 22-year-old black man who lives on the South Side of Chicago - received an unannounced visit by a Chicago Police Department commander to warn him not to commit any further crimes. It turns out that McDaniel was one of approximately 400 people to have been placed on Chicago Police Department's "heat list". These individuals had all been forecast to be potentially involved in violent crime, based on an analysis of geographic location and arrest data -predictive policing: "the application of analytical techniques - particularly quantitative techniques - to identify likely targets for police intervention and prevent crime or solve past crimes by making statistical predictions" - synthetic population is a demographically accurate individual-level representation of a real population -Machine learning algorithms of the kind predictive policing software relies upon are designed to learn and reproduce patterns in the data they are given, regardless of whether the data represents what the model's creators believe or intend. - The algorithms were behaving exactly as expected - they reproduced the patterns in the data used to train them. -To make matters worse, the presence of bias in the initial training data can be further compounded as police departments use biased predictions to make tactical policing decisions. - PredPol, one of the largest vendors of predictive policing systems in the USA and one of the few companies to publicly release its algorithm in a peer-reviewed journal. It has been described by its founders as a parsimonious race-neutral system that uses "only three data points in making predictions: past type of crime, place of crime and time of crime. - Although predictive policing is simply reproducing and magnifying the same biases the police have historically held, filtering this decision-making process through sophisticated software that few people understand lends unwarranted legitimacy to biased policing strategies. The impact of poor data on analysis and prediction is not a new concern.

"What is machine learning" video

-Machine learning: solely focused on writing software that can learn from past experiences -More closely related to statistics and data mining than it is to AI -Machine learning (textbook definition): If a computer program can improve how it performs a certain task based on past experience then you can say it learned. OR The extraction of knowledge from data -Classification: the process whereby a machine can recognize and categorize things from a data set including from visual data and measurement data -Prediction: where a machine can guess predict the value of something based on previous values -Machine learning 3 categories: supervised, unsupervised and reinforcement learning -Supervised learning: where you teach, train the machine using data which is well labeled, the greater the data set the more it learns. Given more unseen data and uses algorithm with past experience to give an outcome. -Unsupervised: machine is trained using data set that doesn't have any labels, the learning algorithm is never told what the data represents -Reinforcement: the training data is unlabeled, however when asked a question about the data the outcome is graded (ex. board game) -Neural net: machine learning technique modeled on the way neurons work in the human brain, given the number of inputs the neuron will propagate a signal depending on how it interprets those inputs -Google using 30 layer neural nets, dropping voice recognition error from 23 to 8%

What are the rules of Human-Robot interaction (video)

-People think that robots are alive, when we interact with them we treat them as if they are alive -Do this in order to make sense of them -We tend to personify robots because Physicality, we respond differently to things in our physical space we treat them like social actors Movement, we automatically project intent on anything that moves. -We respond to the social cues from lifelike machines -Privacy and security is an issue -Companies could exploit emotional connection to get money out of people who use the robots -Robot seal is a very bad replacement for human care, not intended to do. Intended to replace animal therapy in context where we cannot use real animals -People will consistently treat robots more like animals than devices -Pleo experiment: no one wanted to hurt the other persons robot dinosaur toy -Hex bug experiment: had people smash them with mallets after observing it. tested to see how did peoples response change if it had a name and story and the other thing was how peoples natural tendency to empathy related to how they related to the robot People with low empathy for others did not care about the hexbug named Frank, those who scored high on empathy hesitated much more particularly with a name and story behind the bug -Study in Japan with mall patrol robot, children who were unsupervised would harass and hit the robot. -Autistic kid developed relationship with siri and taught the kid new things and articulate his words better. However the programs aren't designed to teach children these things. -Amazon virtual assistant "alexa" doesn't require polite statements, making people more blunt. Could impact children's behavior patterns.

"Utterly horrifying: ex-Facebook insider says covert data harvesting was routine" -Lewis

-Sandy Parakilas, the platform operations manager at Facebook responsible for policing data breaches by third-party software developers between 2011 and 2012, told the Guardian he warned senior executives at the company that its lax approach to data protection risked a major breach. -Once the data left Facebook servers there was not any control, and there was no insight into what was going on." -Facebook's previous policy of allowing developers to access the personal data of friends of people who used apps on the platform, without the knowledge or express consent of those friends. -Facebook took a 30% cut of payments made through apps, but in return enabled their creators to have access to Facebook user data. -"It was well understood in the company that that presented a risk," he said. "Facebook was giving data of people who had not authorized the app themselves, and was relying on terms of service and settings that people didn't read or understand."

Automating Inequality Chapter 3

-Skid row became known as the Poor Man's district in 1930 due to swelled population of migrant workers, grew in diversity tremendously -Housing act of 1949 demolished 7,000+ units of housing, 65,000 building code violations hit Skid Row and when federally funded low-income housing was proposed as replacement middle class resisted -In 1960's Centropolis plan knocked down 7,500 more houses in favor of light industry around the neighborhood as well as money in nearby business district -Blue Book Proposal: protected remaining single room occupancy hotels on Skid Row and encouraged city government and local nonprofits to commit resources to improving housing and social services in the area -Coordinated Entry system created to address mismatch between housing supply and demand in LA county -Coordinated Entry is based on two philosophies: Prioritization which differentiates between two kinds of homelessness: crisis and chronic. Crisis homeless often self-correct and after short stays in shelters find family. Chronic tend to be homeless longer or more often, making permanent supportive housing effective for them. The other is housing first philosophy. Sobriety, treatment compliance, employment are gateway steps introduced to homeless people to help them become housing ready and eventually get them homes and offer support services where appropriate. -VI-SPDAT: Vulnerability Index Service Prioritization Decision Assistance Tool consists of personal information as well as intimate questions the applicant must answer Scores from 1 (low risk of dying or ending up in hospital) to 17 (most vulnerable) are used accordingly. 0-3 don't need housing intervention 4-7 Qualify for limited-term rental subsidies and rapid re-housing 8-17 are assessed for permanent supportive housing Scores paired with vacancy forms and if match is made the person is assigned housing navigator. -Skid row has been ground zero for coordinated entry efforts in LA due to largest number of homeless individuals in 2017 of over 15,000 -Pathways follows a harm reduction/ housing first philosophy and is officially a 90- day shelter despite finding housing for these people in 3 months is nearly impossible -High VI-SPDAT score in LA is a catch 22 due to little permanent supportive housing in the area -Lavan ruling barred city employees from seizing property unless it presents a threat to the public or is evidence of a crime and requires any property collected as "abandoned" to be held secure for 90 days -Out of 57,000 unhoused in LA, 31,000 taken VI test, just over 9,000 have found housing or housing related resources

Robots and Privacy (Calo)

-The clearest way robots implicate privacy is the direct surveillance issue -Another way is that they implicate privacy with introduced new points of access to historically protected spaces -Social nature of robots may lead to new types of highly sensitive personal information implicating what might be called setting privacy. -Surveillance, access, and social meaning are interrelated -The replacement of human staff with robots also presents novel opportunities for data collection by mediating commercial transactions. -Use of a robot capable of connecting to internet within the home creates the possibility for unprecedented access to the interior of the house by law enforcement and hackers alike -Government will be able to secure a warrant for a recorded information within sufficient legal process and gaining live access to the robot and stream of sensory data. -Current technology also offers practical means for individuals to gain access to and even control robots in the home. -Robots are increasingly designed to interact more socially, especially for older people

There will never be an age of artificial intimacy (Turkle)

-These robots can perform empathy in a conversation about your friend, your mother, your child or your lover, but they have no experience of any of these relationships. Machines have not known the arc of a human life. -The coincidence is too convenient: Children will lose the ability to have empathy if they relate too consistently with objects that cannot form emphatic ties. -There are not enough people to care for the elderly; naturally, robots will do that job. There is an epidemic of loneliness; robots will make this a thing of the past -Technologists presented us with artificial intelligence, and in the end it made us look differently, and more critically, at the kind of intelligence that only people have.

"Machine Intelligence Makes Human Morals more important" Video

-We do not know what the machines are learning when we feed them data in machine learning, we just know if it knows it or not - Currently, computational systems can infer all sorts of things about you from your digital crumbs, even if you have not disclosed those things. They can infer your sexual orientation, your personality traits, your political leanings. They have predictive power with high levels of accuracy. Remember -- for things you haven't even disclosed. This is inference. -these systems are often trained on data generated by our actions, human imprints. Well, they could just be reflecting our biases, and these systems could be picking up on our biases and amplifying them and showing them back to us -Our machine intelligence can fail in ways that don't fit error patterns of humans, in ways we won't expect and be prepared for. It'd be lousy not to get a job one is qualified for, but it would triple suck if it was because of stack overflow in some subroutine. - We need to cultivate algorithm suspicion, scrutiny and investigation. We need to make sure we have algorithmic accountability, auditing and meaningful transparency. We need to accept that bringing math and computation to messy, value-laden human affairs does not bring objectivity; rather, the complexity of human affairs invades the algorithms

Secret algorithms threaten the rule of law (Pasquale)

-When an algorithmic scoring process is kept secret, it is impossible to challenge key aspects of it -These assessments are an extension of a trend toward actuarial prediction instruments for recidivism risk. They may seem scientific, an injection of computational rationality into a criminal justice system riddled with discrimination and inefficiency. -In Loomis v. Wisconsin, a judge rejected a plea deal and sentenced a defendant (Loomis) to a harsher punishment in part because a COMPAS risk score deemed him of higher than average risk of recidivating. Loomis appealed the sentence, arguing that neither he nor the judge could examine the formula for the risk assessment -Two forms of automation bias also menace the right to a meaningful appeal. Judges are all too likely to assume that quantitative methods are superior to ordinary verbal reasoning, and to reduce the task at hand (sentencing) to an application of the quantitative data available about recidivism risk -Sonja Starr has argued that what is really critical in the sentencing context is not just recidivism in itself, but the difference a longer prison term will make to the likelihood a convict will reoffend. Algorithmic risk assessment may eventually become very good at predicting reoffense, but what about a risk assessment of risk assessment itself—that is, the danger that a longer sentence for a "high risk" offender may become a self-fulfilling prophecy -Companies are marketing analytics to predict not only the likelihood of criminal recidivism, but also the chances that any given person will be mentally ill, a bad employee, a failing student, a criminal, or a terrorist.

Machine Bias (Propublica)

-Yet something odd happened when Borden and Prater were booked into jail: A computer program spat out a score predicting the likelihood of each committing a future crime. Borden — who is black — was rated a high risk. Prater — who is white — was rated a low risk. -The formula was particularly likely to falsely flag black defendants as future criminals, wrongly labeling them this way at almost twice the rate as white defendants. -White defendants were mislabeled as low risk more often than black defendants.

Amazon Echo Recorded and Sent couple's conversation all without their knowledge (NPR)

-their Amazon Echo was listening and recording their discussion. The device then sent the recording to someone in their contacts — without the couple's knowledge. -The employee sent the couple the sound file that the Echo had sent to him, and they were shocked to realize they had essentially been bugged -"Amazon's Echo uses seven microphones and noise-canceling tech to listen out for its wake word," Washington Post technology columnist Geoffrey Fowler explains. "Doing so, it records about a second of ambient sound on the device, which it constantly discards and replaces. But once it thinks it hears its wake word, the Echo's blue light ring activates and it begins sending a recording of what it hears to Amazon's computers." -"These are potential surveillance devices, and we have invited them further and further into our lives without examining how that could go wrong. And I think we are starting to see examples of that."

Model

...an abstract representation of some process, be it a baseball game, an oil company's supply chain, a foreign government's actions, or a movie theater's attendance

Critical Questions for Big Data

1. Big data changes the definition of knowledge 2. Claims to objectivity and accuracy are misleading 3. Bigger data are not always better data 4. Taken out of context, Big Data loses its meaning 5. Just because it's accessible does not make it ethical 6. Limited access to Big Data creates new digital divides

Example of deontological ethics: Categorical Imperative

An action is morally right if it is in accordance with some list of duties and obligations. •Some actions are always wrong, no matter what the consequence. •Act only according to that maxim that it should become a universal law. •Respect people as valuable in themselves, respect their dignity as persons. Act in such a way that you treat humanity, whether in your own person or in the person of any other, never merely as a means to an end.

Product Perspective

Characteristics of data, such as volume, velocity, variety

Teleological Theory

Consequentialism Worth of an action judged by the consequences End justifies the means Supports paternalistic behaviour if no harm is done

Cambridge Analytica and Facebook Simplified

Facebook exposed data on up to 87 million Facebook users to a researcher who worked at Cambridge Analytica, which worked for the Trump campaign. •Cambridge Analytica received the data from an academic researcher, Aleksandr Kogan, a University of Cambridge psychologist. Kogan built a quiz that was a FB app that harvested data from those who took the quiz AND their friends. •Kogan had access to the data as an academic researcher. But he gave it to Cambridge Analytica, a private company. This was in violation of FB policy, but the company exerted minimal oversight. •Cambridge Analytica wanted the user data to create psychological profiles of users and match personality traits and voter rolls. •FB allowed a third party developer to engineer an app designed to gather user data. FB learned that Kogan had given the data to Cambridge Analytica in 2015 and asked CA to remove the data. FB did not follow up to make sure the company had actually complied. (They did not).

Housing First Philosophy

If you prioritize securing housing for people who are vulnerable to mental illness, addition, disability and so on, then they will be more likely to be able to successfully work on goals such as sobriety and coping with mental illness

Cognitive Perspective

Reveals hidden patterns in the data

Deontological Theory of Ethics

Rule based, universal

Big Data

the process of applying serious computing power—the latest in machine learning and artificial intelligence—to seriously massive and often highly complex sets of information.

Example of Teleological Theory: Utilitarianism

•The ethical action is the one that provides the most good and does the least harm—that is, produces the greatest balance of good over harm. •The outcome of the behavior matters, not the intention. •Actions, rules, and policies are good because they bring about good consequences. •Often paraphrased as "the greatest good for the greatest number." •Two kinds of utilitarianism: Act-based and rule-based.

Algorithm

a computerized analysis that makes use of a mathematical model