Go With the Flow: Effects of Transparency and User Control on Targeted Advertising Using Flow Charts

PowerPoint-presentatie

Go With the Flow: Effects of Transparency and User Control on Targeted Advertising Using Flow ChartsYucheng Jin, Karsten Seipp, Erik Duval, Kartrien VerbertAugment groupHCI @ KU Leuven8 June 2016

0

http://www.gazmuth2.com/wp-content/uploads/2015/06/Denver-City-Council-No-Advertising-Ban-Please.jpg1

Most of people hate seeing ads1

http://zeendo.com/info/wp-content/uploads/2013/02/ch1.png

2

Fear non-consensual use of their data by third partiesFeel irritated by the same flight ad being shown repeatedly2

3

How can we represent the trade-off between value presented by user data (for instance for the advertiser) and value realized through personalization (for instance of relevant advertisements for the user)?

H ow can we deal with such a trade-off?One way to achieve this is to experiment with an approach where the user can control what kind of information he shares with which advertiser, and what he expects in return via personalization.

3

http://www.xmlgrrl.com/blog/2008/05/11/practical-human-centering-and-vrm/http://cyber.law.harvard.edu/projectvrm/Main_Page

4manage relationships with organizationsshare data selectivelycontrol how their data is used(Searls, 2006)

Appear in computer world magazine in 2000In 2006, the projectVRM was born at Harvard Uni.Provide tools for individuals to manage relationships with organizations.Give individuals the ability to share data selectively, without disclosing more personal information than the individual allows.Give individuals the ability to control how their data is used by others, and for how long.

4

5Related work

5

6importance and potential benefits of transparency (TR) and user control (UC) for targeted ads.

K. O'Donnell and H. Cramer. People's perceptions of personalized ads. In Proc. WWW '15 Companion, pages 1293-1298. WWW Steering Committee, 2015.

B. Ur, P. G. Leon, L. F. Cranor, R. Shay, et al. Smart, useful, scary, creepy: Perceptions of online behavioral advertising. In Proc. SOUPS '12, pages 4:1-4:15. ACM, 2012.

L. F. Cranor. Can users control online behavioral advertising effectively? Security & Privacy, IEEE,10(2):93-96, 2012.

Transparency and user control

Several past studies on user perception of targeted advertisinghave discussed the importance and benefits of having TR or UC

no comprehensive research exists investigating their capability to improve targeted advertising.

6

https://www.facebook.com/business/products/ads7

7

https://www.facebook.com/business/products/ads8

text-based explanations, no interactive visualizations to implement TR and UC

8

Positive effects of transparency facilities on trust, agreement, satisfaction and acceptance of E-Commerce recommendations.

(Gregor, 1999; Wang, 2007)

9

9

Visualizing recommender systems

Talk Explorer (Verbert, 2013)10

understand the rationale behind recommendationsto fine-tune various recommendations parameters according to their preferences and needsof other users, tags and suggestions of recommender agents in order to find relevant itemsset visualization

10

TasteWeights (Bostandjiev,2012)11

increase accuracy and recall in recommendations as well as enhanced user experience.

11

12System Design

12

13simplevisual representationintuitive

Simple, visual representation, and intuitive13

14

FLINT (Crews, 1998)RetroGuide (Huser, 2010)Applications of flowchart

examine the utility of flowcharts for various purposes

RetroGuide a flowchart-based analytical framework forquery tasks using a step-based approach14

TransparencyUser control15

16

17

User profileFlow of selectionControl panel

flow chart based visualization to show the process of ad selection and a control panel to configure the user modelNo opt out

17

18Evaluation

We designed a new visualization based on flow chart to support Transparency (TR) and User control (UC) of targeted ads.

We hypothesize that quality and effectiveness of ads can be increased by empowering users to explore and steer the selection process.

18

We conducted a between-subjects study on Amazon Mechanical Turk (MTurk).

200 subjects$1 for each studyaverage time 11 minutes.

We created four experimental conditions:Condition 1 (C1): (No-TR & No-UC) base conditionCondition 2 (C2): (TR & No-UC).Condition 3 (C3): (No-TR & UC).Condition 4 (C4): (TR & UC)19

19

20

C1: (No-TR & No-UC) base condition

C2: (TR & No-UC) C3: (No-TR & UC) C4: (TR & UC)

Subjects

~80% subjects noticed online targeted ads.~10% subjects configured targeted ads. 21

Pu, Pearl, Li Chen, and Rong Hu. "A user-centric evaluation framework for recommender systems."Proceedings of the fifth ACM conference on Recommender systems. ACM, 2011.Materials

We used ResQue and tailored the questionnaire to evaluate four aspects of targeted advertisement:

Quality Behavioral intentionUnderstandingAttitude

Log file

22

user-centric evaluation framework of recommender system 22

5-point Likert scale, Strongly agree - Strongly disagreeQualityBehavioral intentionUnderstanding Attitude23

23

Evaluation stepsIntroduce web app to subjectsLog in to the app with their Facebook accounts. During the trailer, subjects can rate the ads and configure ads if they wish. After the trailer, subjects were asked to complete the questionnaire.24

Playback controls were disabled Que only displayed after the trailerto ensure that subjects were exposed to ads

24

25Results and discussion

Data was not normally distributed (Shapiro-Wilk, df=50, p C1Interest match: TR & UC Context match: limitedAttractiveness: limitedAnnoyance: TR / UC, TR & UC

26

Behavioral intention27

(H=11.42, df=3, p=.01)C4 > C1(H=11.74, df=3, p=.008)C4 > C1; C2 > C1Willingness to click: TR & UC (Log file: 61% subjects click the ads)Willingness to purchase: limited (personal and motivational aspects)Willingness to see: TR , TR & UC

27

Understanding28

(H=13.68, df=3, p=.003)C4 > C1; C3 > C1

Understanding: TR / UC, TR & UC

28

Quality29

Satisfaction: limited (privacy) Confidence: limitedTrust: limited (company credibility and company trust)P42 said that personalizedads make me feel like spying or a violation of my privacy.R. E. Goldsmith, B. A. Laerty, and S. J. Newell. The impact of corporate credibility and celebrity credibility on consumer reaction to advertisements and brands. Journal of Advertising, 29(3):43{54, 2000

complex and context-dependentpart of a process that is not affected by ad-hoc control and insight.29

30Conclusion

30

first implementation of flow charts for targeted adsnew insights in TR and UC for adsProviding only TR improves a user's Behavioral IntentionProviding only UC improves a user's Understanding of the ad selection processProviding both TR and UC improves the aspects Quality, Behavioral Intention, and UnderstandingAttitude does not appear to be affected by either approach.31

Illustrate cause and effect of user traits and preferences on ad selection. from other domains may also be applicable to that of adextending the validity and scope of previous findings31

Limitation

Studies conducted via MTurk may suffer from inattentive or gaming" users.

A small size of data set of ads. 70 elements of 7 ad categories.

A. Kittur, E. H. Chi, and B. Suh. Crowdsourcing user studies with mechanical turk. In Proc. CHI '08, pages 453{456. ACM, 2008.32

32

Thank you for your attention. Yucheng [email protected]?

IWT (IWT-SBO-Nr. 110067).