Ethics of Algorithms: Overview
3
Ethics of Algorithms: Overview Introduction Algorithms—crucial to the field of computer science since the advent of computing—took on even greater importance during the first decades of the twenty-first century as technology companies explored new applications for algorithms within subfields such as artificial intelligence and machine learning. During the 2000s and 2010s, algorithms came to be used for a wide range of purposes, including serving social media users with relevant content, analyzing and identifying images of human faces, predicting criminal activity, and assessing job applicants. At the same time, however, a debate arose regarding how best to manage the ethical implications of algorithmic technologies, which at times raised moral quandaries and made questionable decisions based on biased data. Some individuals and organizations concerned about those ethical ramifications have argued that technology companies and their programmers should take the lead in promoting ethical algorithm design and use and in combatting bias through means such as human-led audits. Some, on the other hand, have asserted that the technology industry cannot regulate itself and that the US government must intervene and regulate algorithms to ensure that they are being designed and implemented in an ethical manner. Understanding the Discussion Algorithm: A sequence of instructions that a computer follows to perform a given task. Artificial intelligence (AI): Computer technology designed to perform intelligence-based tasks, including analyzing data and making decisions, that have historically been carried out by humans. Facial recognition technology: Computer technology designed to detect human faces and, in many cases, match those faces with images of faces stored in a database for the purpose of determining or confirming the identities of the individuals in question. Machine learning: A phenomenon in which computer algorithms learn from the data they process and change accordingly, without direct instruction by human programmers. Predictive policing: The use of various means, including algorithm-based technologies, to predict crimes and prevent them from occurring. Quote from “Ethical algorithm design should guide technology regulation.” The Brookings Institution Artificial Intelligence and Emerging Technology (AIET) report by Michael Kearns and Aaron Roth, January 13, 2020. Photo courtesy Pixabay. History The term algorithm originated in mathematics and was subsequently brought into widespread use within the field of computer science. It refers to a sequence of instructions a computer follows to perform a specific task. The concept of computer algorithms predates the physical creation of computers in the modern sense, arising as early as the mid-nineteenth century in the form of early programs written for proposed —though not yet built—computing devices. Perhaps the most famous of those early algorithms was a program written by the English mathematician Ada Lovelace, created for use with the Analytical Engine computing device proposed by inventor Charles Babbage. Lovelace’s algorithm, published in 1843 as a note in her annotated translation of an Italian paper on the Analytical Engine, was designed to calculate the numbers known as Bernoulli numbers. As computer technology developed from theoretical to physical, algorithms designed for computing devices became increasingly complex, rendering computers capable of performing a variety of Copyright © EBSCO Information Services, Inc. • All Rights Reserved Page 1
Transcript of Ethics of Algorithms: Overview
Ethics of Algorithms: Overview
IntroductionAlgorithms—crucial to the field of computer science since theadvent of computing—took on even greater importance duringthe first decades of the twenty-first century as technologycompanies explored new applications for algorithms withinsubfields such as artificial intelligence and machine learning.During the 2000s and 2010s, algorithms came to be used fora wide range of purposes, including serving social media userswith relevant content, analyzing and identifying images of humanfaces, predicting criminal activity, and assessing job applicants.
At the same time, however, a debate arose regardinghow best to manage the ethical implications of algorithmictechnologies, which at times raised moral quandaries and madequestionable decisions based on biased data. Some individualsand organizations concerned about those ethical ramificationshave argued that technology companies and their programmersshould take the lead in promoting ethical algorithm design anduse and in combatting bias through means such as human-ledaudits. Some, on the other hand, have asserted that the technologyindustry cannot regulate itself and that the US government mustintervene and regulate algorithms to ensure that they are beingdesigned and implemented in an ethical manner.
Understanding the DiscussionAlgorithm: A sequence of instructions that a computer followsto perform a given task.
Artificial intelligence (AI): Computer technology designed toperform intelligence-based tasks, including analyzing data andmaking decisions, that have historically been carried out byhumans.
Facial recognition technology: Computer technology designedto detect human faces and, in many cases, match those faceswith images of faces stored in a database for the purpose ofdetermining or confirming the identities of the individuals inquestion.
Machine learning: A phenomenon in which computeralgorithms learn from the data they process and changeaccordingly, without direct instruction by human programmers.
Predictive policing: The use of various means, includingalgorithm-based technologies, to predict crimes and prevent themfrom occurring.
Quote from “Ethical algorithm design should guidetechnology regulation.” The Brookings Institution ArtificialIntelligence and Emerging Technology (AIET) report byMichael Kearns and Aaron Roth, January 13, 2020. Photocourtesy Pixabay.
HistoryThe term algorithm originated in mathematics and wassubsequently brought into widespread use within the field ofcomputer science. It refers to a sequence of instructions acomputer follows to perform a specific task. The concept ofcomputer algorithms predates the physical creation of computersin the modern sense, arising as early as the mid-nineteenthcentury in the form of early programs written for proposed—though not yet built—computing devices. Perhaps the mostfamous of those early algorithms was a program written bythe English mathematician Ada Lovelace, created for use withthe Analytical Engine computing device proposed by inventorCharles Babbage. Lovelace’s algorithm, published in 1843 asa note in her annotated translation of an Italian paper on theAnalytical Engine, was designed to calculate the numbers knownas Bernoulli numbers.
As computer technology developed from theoretical to physical,algorithms designed for computing devices became increasinglycomplex, rendering computers capable of performing a variety of
Copyright © EBSCO Information Services, Inc. • All Rights Reserved Page 1
complicated tasks. One of the first recorded uses of algorithmswith computers was by Alan Turing, a mathematician best knownfor breaking the Enigma code during World War II who publisheda computational theory to explain the patterns seen in nature.Turing’s goal to capture general processes led to the creationof the Turing Machine, which in turn led the way for generalpurpose computers. Those machines were the first computers ableto execute arbitrary sets of instructions. Following those earlysuccesses, computer scientists made significant strides in theareas of computer technology and artificial intelligence (AI)—algorithm-based computer technology that performs intelligence-based tasks, such as analyzing information and making decisions,that have historically been carried out within human minds. SomeAI technologies function based on specific instructions codedby programmers, while others fall within the realm of machinelearning, in which algorithms process large sets of data, learnfrom that data, and adapt accordingly without further humaninstruction.
During the early twenty-first century, algorithms came to playcrucial roles in both programs and services used regularlyby the public and in more experimental technologies underdevelopment. In the 2000s, social media services, such asFacebook and Twitter, began to use algorithms to determinewhich posts would appear in a user’s newsfeed and in what order,as well as to identify trending content. Algorithms were alsocreated to supply users of a content-providing service, such as amusic- or video-streaming application, with relevant content. Themusic-streaming service Spotify, for instance, uses algorithms torecommend new music to a user based on that individual’s pastlistening habits, while the video-streaming website YouTube usesalgorithms to recommend videos related to one that the user hasjust watched.
Another common use of algorithms is the analysis and taggingof images. The service Google Photos, for example, could usealgorithms to determine that a specific image was a photographof a cat and could then tag the photograph accordingly. Image-analyzing algorithms were also key to the development of facial-recognition technology, which can locate a human face in animage, analyze its features, and compare that face against imagesof faces in a central database to find a match. Such technologycame into regular use both as a personal security feature, enablingusers to unlock their mobile devices with their faces rather thanpasscodes or fingerprints, and as a form of surveillance andpolicing technology, allowing authorities to identify individualsof interest based on sources such as security camera footage.
In addition to using facial-recognition technology, lawenforcement and corrections officials in some areas experimentedwith a practice known as predictive policing, in which algorithmswere used to predict the areas in which crimes were most likely tooccur, which individuals were likely to commit crimes, and whichimprisoned individuals seeking parole were likely to commitfurther crimes if released.
Some technological innovations based on AI and machinetechnologies earned praise and demonstrated the potential ofalgorithms to increase the efficiency, accuracy, and profitability
of certain processes for companies and convenience or betterquality of life for users. However, their emergence into themainstream likewise called attention to the ethical drawbacksof some such technologies. Content-serving algorithms, suchas those used by Facebook and Twitter, drew criticism forexerting undue control over users’ newsfeeds and thereby furtherpolarizing the public politically, and Facebook’s algorithm tohighlight trending news stories struggled to discern factualarticles from so-called fake news. Facial recognition programsmade more errors when trying to identify women andpeople of color than they did when identifying White men,a phenomenon that prompted researchers to investigate thediversity of the datasets used when training such programs andto speak out against the use of facial recognition technologyin policing, as flawed programs could produce false matches.Predictive policing also drew criticism, with some scholarsand advocacy groups asserting that the algorithmic approach tolaw enforcement was based on flawed data, disproportionatelytargeted Black people and residents of poor neighborhoods, andultimately perpetuated systemic racism. Additional technologiesthat presented ethical challenges included an experimentalrecruiting algorithm tested by the retailer Amazon, whichpenalized applications that included the word women’s. Ethicalproblems also arose around the use of algorithms in medicine andfinance.
Ethics of Algorithms TodayIn light of growing concerns regarding the ethical implicationsof AI, machine learning, and other algorithm-based technologies,the ethics and implicit biases of algorithms became the subjectof extensive research throughout the United States, including atcolleges and universities such as the Massachusetts Institute ofTechnology (MIT), Harvard University, and Stanford University.Researchers affiliated with such institutions not only looked intothe causes and ramifications of algorithmic bias, but also exploredmeans of mitigating such biases and tackling the numerousother ethical quandaries the continuously developing AI industryraised.
In response to the increased emphasis on ethics, a number ofmajor technology companies developed boards or committeesdedicated to ethics, particularly regarding their AI research.Microsoft founded its AI and Ethics in Engineering and Research(AETHER) Committee in 2017. Google launched an AdvancedTechnology External Advisory Council (ATEAC) to oversee itsAI work in March 2019 but disbanded the council less than twoweeks later. Social media companies also began ethics-orientedself-regulation initiatives, with Facebook selecting the first groupof members of its Oversight Board, a group focused primarily oncontent-moderation disputes, in 2020.
Although scholars, organizations, and industry leaders with adiverse array of perspectives widely agreed that the use ofalgorithm-driven technology raised serious ethical concerns,opinions regarding the best means of addressing those concernswere split. Many individuals and trade organizations within thetechnology industry argued that technology companies should
Copyright © EBSCO Information Services, Inc. • All Rights Reserved Page 2
take the lead in addressing the issue and could regulate theirown products and practices. In addition to recommending thatcompanies perform regular audits to detect bias or other ethicalissues in their technologies, some proponents of industry-ledreform suggested that the technology industry could also benefitfrom greater diversity in staffing, which might help mitigateimplicit racial or gender bias. By contrast, some experts arguedthat industry self-regulation was ineffective and that existingmeans of government oversight were incapable of keeping upwith the ever-changing technological landscape. Such individualsargued that the US government should introduce new regulationsto protect the public from the misuse of algorithms and toensure that further developments in AI, machine learning, facialrecognition technology, and related areas would be implementedethically.
These essays and any opinions, information, or representationscontained therein are the creation of the particular author anddo not necessarily reflect the opinion of EBSCO InformationServices.
About the AuthorJoy Crelin is a freelance writer and editor based in Wethersfield,Connecticut. She holds a bachelor of fine arts in writing,literature, and publishing from Emerson College.
Bibliography
Brogan, Jacob. “What’s the Deal with Algorithms?” Slate,2 Feb. 2016, slate.com/technology/2016/02/whats-the-deal-with-algorithms.html. Accessed 22 Feb. 2021.
DeAngelis, Stephen F. “Artificial Intelligence: HowAlgorithms Make Systems Smart.” Wired, Sept.2014, www.wired.com/insights/2014/09/artificial-intelligence-algorithms-2. Accessed 22 Feb. 2021.
Heaven, Will Douglas. “Predictive Policing AlgorithmsAre Racist. They Need to Be Dismantled.”MIT Technology Review, 17 July 2020,www.technologyreview.com/2020/07/17/1005396/predictive-policing-algorithms-racist-dismantled-machine-learning-bias-criminal-justice. Accessed 22Feb. 2021.
Kearns, Michael. “How to Build an Ethical Algorithm.”Interview by Dan Costa. PCMag, 27 Feb. 2020,www.pcmag.com/news/how-to-build-an-ethical-algorithm. Accessed 22 Feb. 2021.
Kearns, Michael, and Aaron Roth. “Ethical AlgorithmDesign Should Guide Technology Regulation.”Brookings, 13 Jan. 2020, www.brookings.edu/research/ethical-algorithm-design-should-guide-technology-regulation. Accessed 22 Feb. 2021.
Kearns, Michael, and Aaron Roth. “Who IsResponsible for Biased and Intrusive Algorithms?”Knowledge@Wharton, University of Pennsylvania, 2Dec. 2019, knowledge.wharton.upenn.edu/article/who-is-responsible-for-biased-and-intrusive-algorithms.Accessed 22 Feb. 2021.
Kowalkiewicz, Marek. “How Did We Get Here? The Storyof Algorithms.” Towards Data Science, 10 Oct. 2019,towardsdatascience.com/how-did-we-get-here-the-story-of-algorithms-9ee186ba2a07. Accessed 19 Mar.2021.
Rainie, Lee, and Janna Anderson. “Code-Dependent: Prosand Cons of the Algorithm Age.” Pew ResearchCenter, 8 Feb. 2017, www.pewresearch.org/internet/2017/02/08/code-dependent-pros-and-cons-of-the-algorithm-age. Accessed 22 Feb. 2021.
Joy Crelin
Copyright © EBSCO Information Services, Inc. • All Rights Reserved Page 3
