Investigating Googles Search Engine: Ethics, Algorithms, and the Machines Built to Read Us (Bloomsbury Studies in Digital Cultures)

results: What is a relevant result? Chapter 5: The real cost of search engines: Digital advertising, linguistic capitalism, and the rise of fake news Notation and examples 1 Understanding Google queries and the problem of intentions Contents ix Methodological challenges of studying search engines Particular considerations for collecting search engine results Variables that matter: Search experiments in 2015, 2017, and 2021The rationale behind focusing on same-sex sexual orientation Queries used Capturing the spread of results from the first page Evaluation method Google's public position on how they provide resultsSummary of 2015 results How do variations in terminology and phrasing alter search results? Unimaginable communities How search results change throughout time: 2015, 2017, and 2021 Longitudinal overview: Official languages in each domain Terminology throughout time: 'Homosexual' versus 'gay' Phrasing throughout time: 'Good' versus 'wrong' Conclusion 5 The real cost of search engines: Digital advertising, linguistic capitalism, and the rise of fake news Introduction The economics of google The context of post-Fordism AdWords: Organic versus sponsored results AdWords: The multilingual linguistic market and an economy of bias Google's institutionalization, data collection, and advertising AdWords in the context of 'The Magic System' AdWords and the general intellect The economic profits of discrimination Private profits and public losses Google's international expansion AdSense and post-Fordism: The cost of Google's billboards AdSense and fake news in the 2016 US presidential election The reciprocal relationship between AdSense and Facebook Conclusion Conclusion: What if search engines were actually built to benefit users? Bibliography Index ILLUSTRATIONS Charts 1 This chart shows the distribution of how often users follow each search result provided by Google, also known as Click Through Rate (CTR) distribution 2-4 These graphs show the results for each language across the three geographical domains of the UK, France, and Morocco 5 A graphical representation of the results for each location combined and divided by language 6-8 Graphical representations of terminology between [homosexual] and [gay] and respective variants in each language, all locations combined, 2015 9 Graphical representation of positive and negative query phrasing using English in the UK 10-12 Searching using the official language in each country: 2015, 2017, and 2021 13-14 Searching in English, French, and Arabic with all three countries combined in 2015 and 2021 15-17 Graphs representing the difference in terminology of search queries written in English and searched in the UK in 2015, 2017, and 2021 18 Graphical representation of the difference in terminology of search queries written in French and searched in France in 2021 19 Graphical representation of the difference in terminology of search queries written in Arabic and searched in Morocco in 2021 20-22 Graphs representing the difference in the phrasing between the terms [wrong] and [good] for English in the UK, French translations in France, and Arabic translations in Morocco in 2015, 2017, and 2021 24 This graph shows my repeat of Débarre's study conducted on 12 May 2021, roughly five years later 25 Taken from Mikolov, Sutskever, and Le 'Learning the Meaning Behind Words' 26 A demonstration of how Autocomplete suggests words for very rare queries 27 Following Autocomplete suggestions resulting in a nonsensical query in the early days of RankBrain 28 Further demonstrations of how Autocomplete suggests words for very rare queries, for which it will have minimal direct search data to draw from 29 Further demonstrations of how Autocomplete suggests words for very rare queries, for which it will have minimal direct search data to draw from 30 Further demonstrations of how Autocomplete suggests words for very rare queries, for which it will have minimal direct search data to draw from 31 A photograph taken by the author 32 The photograph of Figure 31 processed by DeepDream 33 Autocomplete results for [b] while on a Google results page for [roller coasters], taken from Effrat et al. 34 Autocomplete results for [b] while on a Google results page for [roller coasters], searched on 6 May 2021 35 Demonstration of personalized search 36 Presentation slide at 13:03 from Marissa Mayer's 'Google Instant Launch Event' detailing the average time taken for the three portions of searching for a query 37 Autocomplete suggestions for [why do bla] 38 An example of a web page result categorized as Pro-gay 'Factual': 'No Ben Carson, Homosexuality Is Not a Choice' 39 An example of a web page result categorized as Pro-gay Opinion: 'Why Being Gay Is Better Than Being Straight' 40 An example of a web page result categorized as Neutral Opinion: debate.o rg 'Is homosexuality wrong' 41 An example of a web page result categorized as Neutral 'Factual': Pew Research Center's article 'The Global Divide on Homosexuality' 42 An example of a web page result categorized as Anti-gay Opinion: Mass Resistance 'What's wrong with being 'gay'? Here's what they don't tell you -and it's really disturbing' 43 An example of a web page result categorized as Anti-gay 'Factual': Christian Answer's 'What's wrong with being gay? Homosexual behavior versus the Bible' 44 'What Does the Bible Say About Homosexuality?' (Creech) 45 The Jehovah's Witnesses page 'Young People Ask: Is Homosexuality Wrong?' that was commonly provided by Google in 2017 as the first result for an overwhelming number of different languages and locations Illustrations xiii 46 Search for the query [Is being homosexual wrong?] using different languages in various countries showing the Google results for 'Is Homosexuality Wrong? | Young People Ask -JW .o rg' and other JW .o rg pages 172 47 AdWord results for 'craft ale' 189 Tables 1 Average terms per query, compiled by the author from various studies. Listed chronologically in order of the 'date of query log collection' 29 2 Autocomplete suggestions collected by Hogan and Luka between 2010-2013 87 3This table shows the complete list of the fifty-four search queries tested and the ratings of the first page of results for each, for the 2015 results 149 I am particularly grateful to Rachel Sykes, Amy Burge, Dorothy Butchard, Rebecca Mitchell and Rex Ferguson for your support and wonderful times spent together. My sincere gratitude goes to my PhD students for our enriching conversations and to all the students who attend the Centre for Digital Cultures events and contribute to an exciting intellectual environment at Birmingham. I am particularly indebted to the founders of Play/Pause and Niall Gallen for your academic generosity and boundless enthusiasm: I look forward to seeing your next exciting steps. Thank you to my undergraduate students, especially those who took Digital Futures and The End of Life, to whom I taught several of the topics in this book. Your curiosity has helped me reflect on the ideas that really matter and your honesty has made me better at explaining them.I am deeply grateful to all the friends who have helped support me during the writing of this book. In particular, thank you to Jack Griffiths and Zenna Taganey for all the pints in the sunshine; Rob Davy-Cripwell for the fun evenings and Slay the Spire comradery, which made the final editing stages bearable; Tristan Potter, Omid Baherli and Charline Jao for the jazz and socialism; Wren for all your support over the years; Greg for your grounded good humour, as well as your tech wizardry that helped Chapter 3 of this book. Special thanks to Ellie and James Tyndall for your love, support and weekends of laughter.A number of people have been instrumental in different parts of my life in enabling me to get to where I am now. I am forever indebted to Julian Kendell. Without your support and generosity in letting me into your classroom, I might never have gone to university. My sincere gratitude to Victor Luftig and everyone I met aboard Semester at Sea, you all changed me in wonderful ways. Thank you to Norroy for the spirit towards education that you passed down. Wesley and Zagreus, you were instrumental in your daily reminders that there is more to life than work.I am particularly grateful to everyone who has directly supported this book, including Anthony Mandal and Jenny Kidd, Ben Doyle, Laura Cope and everyone at Bloomsbury. My deepest appreciation goes to the anonymous reviewers of the first draft of this manuscript. Your generous suggestions thoroughly improved this book, and I am sincerely grateful for the way you gave your feedback. 1 See 'Search engine market share Worldwide, June 2022. ' 2 Google pays a yearly fee to various companies as part of a deal to keep Google Search the default search engine on their devices. For example, industry experts estimate that Google paid Apple $12 billion in 2019 to remain the default on iOS the operating system that powers all of their products. See Duffy (2020). 3 Globally, the current smartphone market share is divided between Google's Android OS (74 per cent) and Apple's iOS (25 per cent). How o en each result is chosen (%) Search Result Rank Click rough Rate (CTR) For Each Search Result on the First Page of Google Results Click rough Rate (CTR) Chitika Insights 2013 study (all devises) Click rough Rate (CTR) Sistrix 2020 study (mobile searches) Chart 1 This chart shows the distribution of how often users follow each search result provided by Google, also known as Click Through Rate (CTR) distribution. It shows that users choose to follow highly ranked results and rarely navigate to Google's second page of results. This chart is based on data from a 2013 Chitika Insights study and a 2020 Sistrix study. Note, Sistrix's study only represents mobile searches, which since 2015 overtook desktop as the most common device used for searching. 9 See Hillis et al. (2012, 41-2) for a more in-depth discussion of studies that support this conclusion. 10 Half Title Series Page Title Page Copyright Page Dedication Contents Illustrations Acknowledgements Introduction: Investigating Google’s Search Engine Google’s dominance The three steps of how search engines work: Crawling, ranking, and query results Step one: Crawling Step two: Ranking Step three: Query results Five key challenges of studying Google’s search engine One: Multiple actors: Search engine optimization and economic incentives Two: Moving targets Three: Each search a partial viewpoint Four: No real alternatives Five: The myth of black boxes Chapter outlines Chapter 1: Understanding Google queries and the problem of intentions Chapter 2: Google’s impact on cognition and memory: Histories, concepts, and technosocial practices Chapter 3: Autocomplete: Stereotypes, biases, and designed discrimination Chapter 4: Google’s search engine results: What is a relevant result? Chapter 5: The real cost of search engines: Digital advertising, linguistic capitalism, and the rise of fake news Notation and examples Chapter 1: Understanding Google Queries and the Problem of Intentions Introduction Categorizing how and what people search The roles of search engines and information retrieval’s question of why Query length and the problems of intention All information is ethical: Searching for [food for snakes] Predicting intentions with a lack of information: Plato, Gadamer, and Derrida Gadamer’s hermeneutics and Plato’s fears of deception Google’s algorithms and Derrida’s monster What kinds of things do people search Google for? Google trends, Brexit, and ‘frantically’ googling after the EU referendum Conclusion Chapter 2: Google’s Impact on Cognition and Memory: Histories, Concepts, and Technosocial Practices Introduction Google’s impact on cognition and memory Kinds of recall from extended minds to transactive memory Technosocial memory practices from oral culture to digital literacy The legacy of naturalized technologies Truth and knowledge for Plato Aristotle’s sensory approach Technosocial memory before Google: The Ars Memoria The science and magic of search Treating the mind as technology: Bacon, Hooke, and modern psychology Conclusion Chapter 3: Autocomplete: Stereotypes, Biases, and Designed Discrimination Introduction The desire for a digital oracle Autocomplete’s minimal academic attention The biases of autocomplete: Stereotypes and discrimination Predicting and shaping user attitudes: The origins of Autocomplete So, how does Autocomplete operate? Second-order stereotyping: Sexist suggestions for female scientists RankBrain and the biases of machine learning Automated misogyny for every individual Speed Speed and judgement: Time to reflect Conclusion Chapter 4: Google’s Search Engine Results: What Is a Relevant Result? Introduction ‘Quantifiable signals’ and Malawian witch doctors What should search engine results be? The Idealists: Search is democratic, relevance can be measured objectively, and answers can exist independently of bias The difficulty with measuring relevance The Contextualists: Search is undemocratic, relevance is a measure of personalization, and all answers are inherently biased Are search results personalized? Methodological challenges of studying search engines Particular considerations for collecting search engine results Variables that matter: Search experiments in 2015, 2017, and 2021 The rationale behind focusing on same-sex sexual orientation Queries used Capturing the spread of results from the first page Evaluation method Google’s public position on how they provide results Summary of 2015 results How do variations in terminology and phrasing alter search results? Unimaginable communities How search results change throughout time: 2015, 2017, and 2021 Longitudinal overview: Official languages in each domain Terminology throughout time: ‘Homosexual’ versus ‘gay’ Phrasing throughout time: ‘Good’ versus ‘wrong’ Conclusion Chapter 5: The Real Cost of Search Engines: Digital Advertising, Linguistic Capitalism, and the Rise of Fake News Introduction The economics of Google The context of post-Fordism AdWords: Organic versus sponsored results AdWords: The multilingual linguistic market and an economy of bias Google’s institutionalization, data collection, and advertising AdWords in the context of ‘The Magic System’ AdWords and the general intellect The economic profits of discrimination Private profits and public losses Google’s international expansion AdSense and post-Fordism: The cost of Google’s billboards AdSense and fake news in the 2016 US presidential election The reciprocal relationship between AdSense and Facebook Conclusion Conclusion: What If Search Engines Were Actually Built to Benefit Users? Bibliography Index "What do search engines do? And what should they do? These questions seem relatively simple but are actually urgent social and ethical issues. The influence of Google's search engine is enormous. It does not only shape how Internet users find pages on the World Wide Web, but how we think as individuals, how we collectively remember the past, and how we communicate with one another. This book explores the impact of search engines within contemporary digital culture, focusing on the social, cultural, and philosophical influence of Google. Using case studies like Google's role in the rise of fake news, instances of sexist and misogynistic Autocomplete suggestions, and search queries relating to LGBTQ+ values, it offers original evidence to intervene practically in existing debates. It also addresses other understudied aspects of Google's influence, including the profound implications of its revenue generation for wider society. In doing this, this important book helps to evaluate the real cost of search engines on an individual and global scale"-- Provided by publisher