Vous êtes sur la page 1sur 4

The Crowd and Persuasion - A Necessity for Individualized Persuasive Technologies?

Pascal Lessel DFKI GmbH Saarbr ucken, Germany pascal.lessel@dfki.de


ABSTRACT

Antonio Kruger DFKI GmbH Saarbr ucken, Germany antonio.krueger@dfki.de


see [4, 16]), and debates and discussions on this quality have taken place (e.g. [3, 6]). 2. In contrast to Wikipedia where the motivations to participate are manifold and are often a research topic (e.g. [1, 12]), in recent years, companies have arisen which utilize crowdsourcing to satisfy customers by simply using money as incentive and motivation. A good example for this is the so called crowd-testing, in which desktop software, mobile apps or websites are tested by dozens or hundreds of people, who earn money per bug found. Crowd-testers submit not only their own bugs but are in general also able to comment on other bugs. In both examples, it is crucial that the members of the crowd are motivated to participate otherwise, the systems will not work. Thus, it is relevant to try to answer the question why people are motivated to participate and whether we are able to increase every individuals participation. The crowd-testing companies1 in the second example have already made use of different persuasive strategies: for example competition (i.e. you receive more testing opportunities if you are better than the other testers), comparison (i.e. a high score list shows an aggregated score and a placement with respect to others) and self-monitoring (i.e. you see your detailed performance over several months, as well as other statistics). This is especially interesting as the individual crowd-tester, as stated, earns money on a per-bug basis; one might argue that this alone is a good motivation. In contrast, such persuasive strategies seem to be currently absent on Wikipedia where people are not paid for participating. Nonetheless, both systems seem to work with great success. This raises two questions, which are directly related to the topic of this workshop: First, do systems that aim to work with a lot of different people need different persuasive technologies? Second, if we integrate persuasive technologies, will it be useful to personalize these technologies on an individual basis?
POSITION STATEMENT

In this position paper, we discuss why we think that personalization on the level of persuasive technologies used is an important step for the next generation of persuasive systems. We focus on crowdsourcing / crowd-based computing in which many people (typically hundreds or thousands) participate to solve a specic goal and in which persuasive technologies that do not follow a one-size-ts-all-approach seem to be most suitable. We continue with a discussion on whether the crowd can also be used to personalize behavior change technology. We close this paper with an overview of recent projects that will serve as a testbed for our investigation of personalizing behavior change systems.
Author Keywords

Persuasion; personalization; crowdsourcing; crowd computing.


ACM Classication Keywords

J.2 Computer Applications: Social and Behavioral Sciences; H.5.m. Information Interfaces and Presentation (e.g. HCI): Miscellaneous
INTRODUCTION

Crowdsourcing is a way to collect knowledge, opinions, decisions and skills to reach a better outcome following the idea of the wisdom of crowds [15]. This idea states that the knowledge of one person is outperformed by the knowledge of many. Nowadays, we can see crowdsourced approaches used online to enhance quality in different areas. Two examples illustrate this: 1. Wikipedia seems to be the obvious example for crowdsourced material. Many people participate and add their knowledge to create and enhance Wikipedia articles. It is important to note that basically everyone can edit every article (often backed up by references or at least comments on why changes are made). Much research concerning the article quality has been conducted already (for instance

By checking the given examples, we might develop an intuition, but answering this for the general case is an interesting research question. In the remainder of this paper, we will focus on the persuasion of the crowd and why we think that a personalization on the level of the persuasive strategies seems important. We will also start a discussion on how the crowd might be used to persuade a single individual with the help of persuasive technologies.
1

Copyright is held by the authors/owners. CHI 2014. Workshop: Personalizing Behavior Change Technologies

Consider for example uTest, Inc: www.utest.com

Why one-size-ts-all seems not to be enough

Considering different publications of recent years, it can be observed that specic behavior change domains are targeted (e.g. sustainability or health) and the researchers often present one persuasive system which is evaluated with a different number of people. Positive effects and results are usually found by the authors, which shows that the overall approach and the persuasive strategies used seem to work as expected. Nevertheless, large variances can often be found on an individual level [9]. The question arises how the result would change if a personalization on the level of the persuasive technologies were implemented. This might provide even better overall results. This belief is supported by some work already done in the area of persuasive technology (e.g. [8]) and (at least to a certain degree) by research done in the area of recommender systems. These systems also belong to the set of persuasive technologies as they try to inuence the users behavior directly. For example, Amazon utilizes a recommender system to change the buying behavior of their customers. The retail area is not the only domain in which recommender systems are used; you can nd them often on the web (e.g. Youtube recommends videos and Facebook suggests friends). A recommender can use different strategies to derive recommendations, such as collaborative ltering and content-based recommendation, as major ones (see [14] for an overview). It seems reasonable that there are individual differences in the way people recognize recommendations: In collaborative ltering, the users behavior is compared to that of other users. Objects are then recommended based on other users decisions that are similar to the querying ones. The contentbased recommendation uses a different approach here, the system derives a set of features based on the users historical behavior. To do that, items are equipped with a set of features. By comparing the users feature vector (populated by his interaction with earlier items) with the products feature vectors, scores are calculated and used to give a recommendation. By comparing these approaches, we have on the one hand a social component in which the recommendation is dependent on the behavior of others, whereas on the other hand, the users behavior is sufcient for the recommendation. Often, the information on how the recommendations are derived is presented to the user (which is called transparency and is generally important for trust in the recommendations [14] )). At least in the area of music recommendation, evidence was found that the perceived usefulness changed depending on the approach used [7]. Nonetheless, it is unclear whether personality traits might play a role in the perception of the recommendations. Assessing these traits and using them in the selection process of the strategy might lead to better suggestions in terms of accuracy and perceived usefulness. In our opinion, similar effects will be observable in other persuasive strategies as well. We see the personalization of persuasive strategies as a crucial next step towards an improvement of persuasive technologies in general. The topics (or domains) usually targeted within persuasive technologies are important and matter for a large group of users. In general, the usual persuasive system targets a crowd of people (e.g.

the group of smokers [11]) and provides them with a system. Often these systems make use of different pieces of persuasive technologies, but a personalization function is not available, as every user is persuaded with the same strategy. If we assume that the aforementioned groups not only consist of 20 people but rather tens of thousands, it seems highly unsuitable that only one specic one-size-ts-all approach might sufce. The view that persuasive technologies target such high numbers is not only a possibility for future systems but is already the case or at least conceptually integrated: Amazon utilizes its persuasive marketing strategies for its customers; mobile sports apps such as Runtastic2 or approaches such as the WaterBot [2], which in theory targets a large user base, are examples of that. Moreover, solving the problem for crowds will most likely also improve the situation for smaller groups or even individuals, as the same personalizing strategies can be applied. If, as the outcome of this workshop and the subsequent research, a framework for personalizing the persuasive technologies becomes available, the examples given in the previous section will also benet, as they are solely designed to work with a large number of users. In general, all these examples underline the importance of the workshop questions.
Using the crowd as a personalized persuasive strategy

When we use persuasive technologies nowadays, it seems often to be the case that we have one computerized system which follows a certain strategy to persuade users of the system. But when we look back in history (back to the Ancient Greeks), we see that persuasion was an important topic of rhetoricians who tried to persuade others through speech. The question arises why we now rely on the persuasive computerized system and only implicitly integrate social aspects (e.g. through competitions or social comparisons). Why do we not use the social component and use the crowd which is supported by computerized systems to persuade individuals as a persuasive technology? This would not only be restricted to an individuals social circle but might also be extended to nd the most suitable (potentially unknown) people in a big user base that might give the best persuasive advice to the user in the given situation. For instance, every individual in the crowd could propose a strategy, followed by a subsequent voting process. Moreover, if we have dedicated ways of providing direct feedback to the crowd on the impact of the chosen strategy and user acceptance, every individual might be able to adapt the proposed strategy, which would lead again to a personalized behavior change technology. Thereby the overall goal would be to combine the advantages of persuasive technologies with the advantages of crowd computing. More specically, we would utilize the human persuasive qualities (compare for example salesmen [10]) not only in the design process of the persuasive technology, but would also use it at runtime. Nevertheless, this might not be an easy task, as many domains demand a long-term consideration (e.g. relapse rate of former smokers) and a crucial question would be: what does the crowd need to know to give good persuasive advice or to adapt the strategies?
2

https://www.runtastic.com/

To give an example of the idea to couple a crowd and a single user, consider the Chorus system [13], in which users are able to ask questions to the system and in which, with the help of a proposal and voting system, the most suitable answers to these questions are derived by the crowd and are provided to the user. By that the user engages in a dialog with the crowd. This system uses certain persuasive elements (at least for the crowd) but might be improved further by the results and progress towards personalized persuasive technologies. We want to stress that we are aware that this social component might not be sufcient for every person depending on their personality traits (similar to the collaborative ltering vs. content-based recommendation example above); nevertheless, if we design our system in such a way that it will also work without this social component, we might reach a certain exibility. Again, it is still an open question to be targeted in the future, how it can be decided which strategy will work and which will not.
Recent projects

CONCLUSION

This workshop targets important aspects that, after targeted and answered, will bring the eld of persuasive computing to the next level. A one-size-ts-all-approach does not seem suitable if systems should persuade many different users, which seems to be the case for the majority of the domains in which persuasion is used nowadays. The rst author is just at the beginning of his PhD and one part of his work will explore the options for personalizing behavior change technologies with the help of the crowd. Participation in this workshop will not only help to acquire an overview on the state of the art, but by shaping a research roadmap for this topic, mutual benet seems reasonable.
REFERENCES

1. Antin, J. My Kind of People?: Perceptions About Wikipedia Contributors and Their Motivations. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, CHI 11, ACM (2011), 34113420. 2. Arroyo, E., Bonanni, L., and Selker, T. Waterbot: Exploring Feedback and Persuasive Techniques at the Sink. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, CHI 05, ACM (2005), 631639. 3. Britannica, E. Fatally Flawed: Refuting the recent study on encyclopedic accuracy by the journal Nature. 2006. 4. De la Calzada, G., and Dekhtyar, A. On Measuring the Quality of Wikipedia Articles. In Proceedings of the 4th Workshop on Information Credibility, WICOW 10, ACM (2010), 1118. 5. Fogg, B. A Behavior Model for Persuasive Design. In Proceedings of the 4th International Conference on Persuasive Technology, Persuasive 09, ACM (2009), 40:140:7. 6. Giles, J. Internet encyclopaedias go head to head. Nature 438, 7070 (December 2005), 900901. 7. Hu, R., and Pu, P. Acceptance Issues of Personality-based Recommender Systems. In Proceedings of the 3rd ACM Conference on Recommender Systems, RecSys 09, ACM (2009), 221224. 8. Kaptein, M., De Ruyter, B., Markopoulos, P., and Aarts, E. Adaptive Persuasive Systems: A Study of Tailored Persuasive Text Messages to Reduce Snacking. ACM Transactions on Interactive Intelligent Systems 2, 2 (June 2012), 10:110:25. 9. Kaptein, M., and Eckles, D. Heterogeneity in the Effects of Online Persuasion. Journal of Interactive Marketing 26, 3 (2012), 176 188. 10. Kaptein, M., Lacroix, J., and Saini, P. Individual Differences in Persuadability in the Health Promotion Domain. In Persuasive Technology, T. Ploug, P. Hasle, and H. Oinas-Kukkonen, Eds., vol. 6137 of Lecture Notes in Computer Science. Springer Berlin Heidelberg, 2010, 94105. 3

Currently, we are working on different projects which will serve us as a testbed for the aforementioned discussion points: 1. Crowd-based creation of outdoor exercises: This project targets people already keen on doing sports and tries to persuade them to do more exercises outdoors. For this, we utilize the concept of virtual tness trails and have two research opportunities: First, how can we persuade the individual user of the app to participate in the crowd-based process to add new outdoor spots at which other users can do exercises? Second, what impact does the crowd subsequently have on the individual user? Can the crowd persuade individual users to do more exercises? 2. A novel persuasive household account book: In the design process of the account book, we consider the behavior model [5] and focus on solutions for the trigger, ability and motivation. Within this project, we create a household account book which is able to automatically derive the expenses through photo-taking of a receipt (ability) and will integrate a GPS- and activity-based reminder functionality (trigger). After validation that the trigger and the reminder functions are working as expected, we will concentrate on the motivation part and will also cope with different personality traits of people using our account book later on. Our research started with a receipt analysis and we learned that we will also need to integrate a crowd-based component to enhance the optical character recognition approach to provide an easy way to add receipt entries. Thus, the results on how to motivate the crowd in the rst project can also be validated in this domain. 3. Augmented cigarette case: Within the area of smoking cessation, we will create an enhanced cigarette case which is always online, uses a display and automatically logs the number of cigarettes smoked. This technology will not only allow self-monitoring but will also allow us to test which options and impact an anonymous crowd can have on the smoker and if eventually the crowd is able to convince the smoker to stop this behavior.

11. Khaled, R., Barr, P., Noble, J., Fischer, R., and Biddle, R. Fine Tuning the Persuasion in Persuasive Games. In Proceedings of the 2nd International Conference on Persuasive Technology, PERSUASIVE07, Springer-Verlag (2007), 3647. 12. Kuznetsov, S. Motivations of Contributors to Wikipedia. SIGCAS Computers & Society 36, 2 (June 2006). 13. Lasecki, W. S., Wesley, R., Nichols, J., Kulkarni, A., Allen, J. F., and Bigham, J. P. Chorus: A Crowd-powered Conversational Assistant. In Proceedings of the 26th Annual ACM Symposium on User Interface Software and Technology, UIST 13, ACM (2013), 151162.

14. Ricci, F., Rokach, L., Shapira, B., and Kantor, P. B. Recommender Systems Handbook, 1st ed. Springer-Verlag New York, Inc., 2010. 15. Surowiecki, J. The Wisdom of Crowds: Why the Many Are Smarter Than the Few and How Collective Wisdom Shapes Business, Economies, Societies and Nations. Doubleday, 2004. 16. Yaari, E., Baruchson-Arbib, S., and Bar-Ilan, J. Information Quality Assessment of Community Generated Content: A User Study of Wikipedia. Journal of Information Science 37, 5 (Oct. 2011), 487498.

Vous aimerez peut-être aussi