東華大學圖書館 |

Measuring the Harms of Personalization Through Advertising.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Measuring the Harms of Personalization Through Advertising./
作者:	Ali, Muhammad.
面頁冊數:	1 online resource (179 pages)
附註:	Source: Dissertations Abstracts International, Volume: 85-05, Section: A.
Contained By:	Dissertations Abstracts International85-05A.
標題:	Computer science. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30811179click for full text (PQDT)
ISBN:	9798380847735

Measuring the Harms of Personalization Through Advertising.
Ali, Muhammad.

Measuring the Harms of Personalization Through Advertising. - 1 online resource (179 pages)

Source: Dissertations Abstracts International, Volume: 85-05, Section: A.

Thesis (Ph.D.)--Northeastern University, 2023.

Includes bibliographical references

Modern online services like social media and content sharing platforms almost always have more content than what a single user can consume. Therefore, they often have to select a subset of content that is shown to each user. To perform this selection, personalization has emerged as a nearly ubiquitous solution, where users are presented only the content that the platform considers most relevant to them. This mechanism can lead to users having varied and unique experiences, focused on content that they find interesting. Personalization is also a useful tool to demote content that might be low quality, or to introduce users to novel content. However, despite increasing engagement, there are concerns that it can lead to adverse user impacts for certain domains of content. Within these domains, judging relevance can harm users by excluding them from potentially life-changing opportunities, such as those for jobs and housing. It can also lead to inclusion of problematic content into users' experiences, such as misinformation and deceptive offers.In this dissertation, I focus on the harms of personalization. I use targeted advertising as a case study of personalization and measure the adverse effects it could have on users. I focus specifically on Facebook's advertising system, the second largest, and one of the most feature-rich advertising platforms in the world. First, I build a series of algorithm auditing methods to isolate the influence of personalization with only black box access to the advertising platform. Second, I employ these methods to measure adverse user outcomes in multiple ad domains: jobs, housing, political campaigns, and problematic ads, as judged by users.Advertising relies heavily on personalization to identify the users who an ad is relevant to. However, unlike other systems, it allows the advertisers to specify an explicit target audience. I leverage this targeting mechanism to design experimental methods that isolate the role of personalization specifically, and control for variance in audience and market competition. Advertising platforms also frequently provide explanations to the users about why an ad was shown to them. I exploit these explanations to measure the impact of personalization in observational data collected from real users. In a series of large scale measurements, I first identify gender and racial disparities in job and housing ads due to personalization, even when the advertiser targets broadly. Second, for election campaign ads, I find evidence of personalization along users' political leaning as well as price differentiation, which potentially limits a candidate's ability to reach non-aligned voters. Third, I investigate individual user experiences of problematic ads, finding that ads considered clickbait and scam by users deliver primarily to a minority of users, who have disparately high exposure to such content. Together, these measurements quantify the extent to which personalization---independent of the advertiser's intent---produces harmful outcomes for users.Through a detailed study of advertising, this dissertation contributes methods to measure the black box of personalization, and provides insights into its adverse effects.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2023

Mode of access: World Wide Web

ISBN: 9798380847735Subjects--Topical Terms:

523869
Computer science.
Subjects--Index Terms:

Human-computer interactionIndex Terms--Genre/Form:

542853
Electronic books.

Measuring the Harms of Personalization Through Advertising.
LDR:04595nmm a2200397K 4500 001 2364637
005 20231130105858.5
006 m o d
007 cr mn ---uuuuu
008 241011s2023 xx obm 000 0 eng d
020 $a 9798380847735
035 $a (MiAaPQ)AAI30811179
035 $a AAI30811179
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Ali, Muhammad. $3 3432704
245 1 0 $a Measuring the Harms of Personalization Through Advertising.
264 0 $c 2023
300 $a 1 online resource (179 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 85-05, Section: A.
500 $a Advisor: Mislove, Alan.
502 $a Thesis (Ph.D.)--Northeastern University, 2023.
504 $a Includes bibliographical references
520 $a Modern online services like social media and content sharing platforms almost always have more content than what a single user can consume. Therefore, they often have to select a subset of content that is shown to each user. To perform this selection, personalization has emerged as a nearly ubiquitous solution, where users are presented only the content that the platform considers most relevant to them. This mechanism can lead to users having varied and unique experiences, focused on content that they find interesting. Personalization is also a useful tool to demote content that might be low quality, or to introduce users to novel content. However, despite increasing engagement, there are concerns that it can lead to adverse user impacts for certain domains of content. Within these domains, judging relevance can harm users by excluding them from potentially life-changing opportunities, such as those for jobs and housing. It can also lead to inclusion of problematic content into users' experiences, such as misinformation and deceptive offers.In this dissertation, I focus on the harms of personalization. I use targeted advertising as a case study of personalization and measure the adverse effects it could have on users. I focus specifically on Facebook's advertising system, the second largest, and one of the most feature-rich advertising platforms in the world. First, I build a series of algorithm auditing methods to isolate the influence of personalization with only black box access to the advertising platform. Second, I employ these methods to measure adverse user outcomes in multiple ad domains: jobs, housing, political campaigns, and problematic ads, as judged by users.Advertising relies heavily on personalization to identify the users who an ad is relevant to. However, unlike other systems, it allows the advertisers to specify an explicit target audience. I leverage this targeting mechanism to design experimental methods that isolate the role of personalization specifically, and control for variance in audience and market competition. Advertising platforms also frequently provide explanations to the users about why an ad was shown to them. I exploit these explanations to measure the impact of personalization in observational data collected from real users. In a series of large scale measurements, I first identify gender and racial disparities in job and housing ads due to personalization, even when the advertiser targets broadly. Second, for election campaign ads, I find evidence of personalization along users' political leaning as well as price differentiation, which potentially limits a candidate's ability to reach non-aligned voters. Third, I investigate individual user experiences of problematic ads, finding that ads considered clickbait and scam by users deliver primarily to a minority of users, who have disparately high exposure to such content. Together, these measurements quantify the extent to which personalization---independent of the advertiser's intent---produces harmful outcomes for users.Through a detailed study of advertising, this dissertation contributes methods to measure the black box of personalization, and provides insights into its adverse effects.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2023
538 $a Mode of access: World Wide Web
650 4 $a Computer science. $3 523869
650 4 $a Web studies. $3 2122754
653 $a Human-computer interaction
653 $a Machine learning
653 $a Personalization
653 $a Platform governance
653 $a Advertising
655 7 $a Electronic books. $2 lcsh $3 542853
690 $a 0984
690 $a 0646
690 $a 0800
710 2 $a ProQuest Information and Learning Co. $3 783688
710 2 $a Northeastern University. $b Computer Science. $3 1678818
773 0 $t Dissertations Abstracts International $g 85-05A.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30811179 $z click for full text (PQDT)