Power Ballads: Deploying AversiveEnergy Feedback in Social Media

AbstractThis paper reports on the pilot evaluation of “PowerBallads”, an evocative social media application whichdisplays aversive feedback based on excessivehousehold energy usage. Work by other researchers inpersuasive technologies has previously suggested theuse of aversive feedback should be avoided as it leadsto a lack of engagement by users. This work evaluateswhether punishment of non-desirable behaviourdiscourages users from engaging with a persuasiveapplication. To this end we recruited 9 households touse the Power Ballads application over a period of 4weeks. We found the use of aversive feedback did notact as a deterrent to regularly interacting with theapplication through evaluating user engagement.

KeywordsEnergy Saving, Persuasive Technology, Sustainability,Social Networks, Aversive Feedback, Motivation

ACM Classification KeywordsH5.m. Information interfaces and presentation (e.g.,HCI): Miscellaneous. J.4. Social and BehaviouralSciences: Psychology

General TermsDesign, Experimentation, Human Factors

Copyright is held by the author/owner(s).

CHI 2011, May 7–12, 2011, Vancouver, BC, Canada.

ACM 978-1-4503-0268-5/11/05.

Derek FosterSchool of Computer Science,University of Lincoln,Lincoln, LN6 7TS, UKdefoster@lincoln.ac.uk

Conor LinehanSchool of Computer Science,University of Lincoln,Lincoln, LN6 7TS, UKclinehan@lincoln.ac.uk

Shaun LawsonSchool of Computer Science,University of Lincoln,Lincoln, LN6 7TS, UKslawson@lincoln.ac.uk

Ben KirmanSchool of Computer Science,University of Lincoln,Lincoln, LN6 7TS, UKbkirman@lincoln.ac.uk

2

IntroductionRecent interest in the use of persuasive technologies[10] has produced a range of interactive interventionsdesigned to change behaviours in areas such as health[3] and sustainability [11][15]. However, very few suchstudies have identified, in detail, the reasoning behindtheir selected behavioural change mechanisms or thecognitive processes that their users are expected toexperience.

Froehlich et al [12] recently undertook a comparativereview of papers generated by both the environmentalpsychology and HCI communities which usedpersuasive technologies and stressed the need for HCIresearchers to collaborate with environmentalpsychologists in order to leverage their expertise whenattempting to change target behaviour. This suggeststhat the persuasive technologies that are beingdeveloped by the HCI community would benefit from afuller understanding of psychological behaviour changestrategies.

Persuasive technologies have typically aimed tomotivate behaviour change primarily throughpresenting positive feedback when the desiredbehaviour is observed [5][2]. Indeed, well knownrecent literature by the HCI community [6][1]specifically recommends that only good behaviour isrewarded, thus offering no consequences when usersdo not meet their goals. Interestingly, this is in sharpcontrast with the findings from applied psychology,where it is suggested that corrective feedback – i.e.presenting consequences regardless of the behaviourobserved – is often the optimum approach [13].Furthermore, our own recent work has discussed how

playful aversive feedback can be harnessed to reduceenergy consumption in the kitchen environment [14].

Those who suggest that aversive feedback is notappropriate for online behaviour-change interventions[6][1] often refer to the problem of users becomingdisengaged when punishment is delivered. It is arguedthat the easiest way for a user to avoid punishment insuch an application is to stop using the application,rather than changing their behaviour. It is important torealise that this is an assumption rather than anempirical finding.

In this work-in-progress we investigate whether playfulaversive stimuli can function as an integral part of anapplication designed to help users monitor and reducetheir domestic energy consumption.

Application DesignWe have devised an application that promotesbehaviour change primarily through playful aversivefeedback. The Power Ballads Facebook application usedin this pilot study is a simple, early prototype. A basicinterface was developed that delivered feedback onparticipant’s comparative energy by time splicing theprevious 48 hours of usage. A commercial off-the-shelfhome energy monitoring system from the companyCurrent Cost - consisting of an ENVI display and Bridge[7] - was used to connect energy data to the PowerBallads application.

When a user visits the Power Ballads application fromtheir Facebook account, the interface informs them oftheir energy usage by means of a large notificationdisplaying ‘Yes!’ if they have saved energy or ‘No!’ ifthey have increased their usage, as shown, for

Figure 1. Power Ballads notificationinterfaces for energy usage

3

example, in figure 1. If the user increases their energyusage then they are effectively punished by theapplication automatically publishing aversive feedbackin the form of a public post to their Facebook newsfeed.This posted message would then be viewable by theirFacebook friends. The post highlighted they were usingmore energy over the past 24 hours compared to theprevious 24 hours, effectively disclosing their highusage, or undesirable behaviour, to their friends.

As energy consumption is an important issue indomestic households, it was apparent that thepublishing of aversive feedback about excessive usageon participants’ Facebook page may be undesirable andcould lead to the type of disengagement discussed in[6][1]. We were interested in exploring whetherpresenting this information in a playful manner mayavoid the problem of disengagement. To this end, wechose to present popular UK chart music as theaversive stimuli posted on participants Facebook wall.Participants were initially screened and accepted basedon their strong disliking for this type of modern chartmusic. An example of a feedback post is shown infigure 2, complete with comments added by the userand their Facebook friends.

ImplementationOur software implementation utilises both the FacebookGraph and Pachube API’s [9][4] to display participantslive online energy data on the Facebook platform. TheCurrent Cost Energy Bridge and Envi display were theselected home energy monitor components for thestudy. The energy bridge provides the capability tosend energy data directly online to the Pachube servicevia a home router. Figure 3 illustrates the technicalimplementation of Power Ballads.

Figure 2. Anonymised example of an aversive feedback post

Pachube is a data brokerage store for the ‘internet ofthings’ and provides an online REST API that allowsauthorised web applications to query sensor data. Thisis an important step forward in the evolution of homeenergy monitoring as most monitors to date require PCsoftware (with powered on PC) to send energy dataonline where it can be used by third party applications.By using cutting edge hardware the process ofmonitoring is more transparent to the end user bysending energy data directly online, increasing the userexperience quality and saving energy.

Experimental Method

AimsThe aims of the study were twofold: 1) To investigatethe effectiveness of aversive feedback to engage usersin eco-feedback applications and 2) Analyse usercomments generated from the aversive feedback postsfor emergent themes that could then be exploited insubsequent eco-feedback applications.

Figure 3. Technical implementation ofPower Ballads

4

ParticipantsUsing a mixture of purposive and conveniencesampling, nine lead participants were recruited: 4 ofwhich were female, from nine households.

DesignEach household was given an energy monitor for aperiod of 5 weeks. At the end of the first weekparticipants were given access to the Power Balladsapplication on Facebook and could then view theirprevious 24 hour energy usage in graph form and viewenergy notifications based on whether or not they aresaving energy in the previous 48 hour period. In orderto view Power Ballads each participant was required tofirst login to Facebook and start the application.Aversive feedback would only be delivered when loggedin giving the opportunity to examine whetherparticipants were engaging with the application orsimply avoiding punishment by not using it. A shortend-of-study qualitative questionnaire was alsocompleted to elicit participants’ subjective experience ofinteracting with Power Ballads.

ResultsDuring the course of the study 5 participants activelyused Power Ballads, 1 participant did not realise theyhad to log into the application to view feedback and 3had compatibility problems with their home routers inrecording their home energy data. Thus, the data fromfive of the nine original participants is presented here.

In total, 167 visits (75% in the first two weeks) weremade to the application with 50 aversive newsfeeditems posted to the participants Facebook wall. Thisequates to around 30% of all visits to the applicationresulting in a punishment post with the remaining 70%

bringing about a notification indicating they weresaving energy. It highlights participants were engagingwith the application by logging in and therefore nottrying to avoid punishment. From the 50 aversivenewsfeed posts an additional total of 57 user-generatedcontent items were created consisting of 41 commentsand 16 ‘Like’ clicks suggesting that playful dialoguetook place. These user generated items were createdby participants and non-participants who wereFacebook friends. Additionally, 5 questionnaires werecompleted from all active users of Power Ballads.

Figure 4 highlights participant engagement throughoutthe study. The graph illustrates the number of times anaversive post was created for each participant and thenumber of times overall they responded via commentsto their own posts. Also shown is the total number ofother feedback items including comments and ‘Like’clicks from other participants and Facebook users. Theresults show that participants engaged with their ownand others’ aversive feedback.

Figure 4. Participant engagement with aversive feedback

5

DiscussionWith the results suggesting that aversive feedback maybe effective in successfully engaging users (or at leastdoes not disengage users), it is worthwhile to look atsome of the user-generated comments that werecreated based on the aversive newsfeed posts.

Engagement on energy usage was present,Participant C: “277...are you sat in the dark?” withParticipant B responding with:” is 277 fairly low? It isnot extremely bright in here, but still can recognisefurniture and Aga!” Similarly, another dialogue,Participant C: “313 average and you have used moretoday? Did you sit in the house with the lights offyesterday?” with Participant D responding: “Yesterdayit was 233 and that was higher than before that! Notsure what's been different today though.”

Interestingly, some of the dialogue was based ondetermining which appliances may be causing theaversive feedback, Participant D: “You're using themost out of Juliet you and me today! Washing day?!"with Participant A responding, “Funnily enoughyeah…the washing machine, caught up on a load ofwashing!! Just shows you”. Participant D respondeddirectly to the aversive feedback when their personalusage peaked at its maximum with “Oh good god! Iguess that's what i get for having the oven on :( ....buti also get home-made bread...not sure it's worth myreputation though!!”

Engagement surrounding the use of chart music asaversive feedback was also present, Non-participant:“jls???” [teenage boy band] with Participant Dresponding: “There needs to be a smiley for hangingyour head in shame!!”. Additionally, the quote, “Oh no,

what are you doing to my reputation!” by ParticipantC was a direct response to an aversive feedback post.

Due to insufficient space we are unable to present afuller breakdown of all the comments. However, themajority of comments were centred on two themes, thefirst based around energy usage and the appliancespotentially responsible for causing the aversivefeedback. The second theme focused around friendlybanter related to the chart music used in the newsfeedposts. These two emergent themes were mirrored inour questionnaire data.

Conclusions and Further WorkThe wider goal of this study was to evaluate whetherpunishment of non-desirable behaviour can function asa useful part of larger future studies. It is apparent thatthis type of feedback does not necessarily lead todisengagement by users if presented carefully. We donot suggest that presenting only aversive feedback isan ideal method of designing a persuasive application;indeed our questionnaire data revealed thatparticipants A and C would have liked supplementaryreward posts when saving energy. However, we havedemonstrated that aversive stimuli do not necessarilybring about disengagement. Rather, as the psychologyliterature suggests that aversive stimuli can function asa valuable component in behaviour changeinterventions. We have demonstrated that it does notnecessarily lead to participant disengagement, andtherefore conclude that it is important that this type offeedback should not be simply ignored when designingpersuasive applications, as has been suggested byother researchers.

6

A common criticism levelled at technology-centricsustainability interventions is that they use nonsustainable resources themselves. Furthermore, manypersuasive technology studies that are designed tomotivate reduction in energy use have been shown tobe successful in the interim. Despite this, more oftenthan not they tend to require feedback to be deliveredpermanently, albeit at a lower frequency over aprolonged duration [8]. It could be argued thattechnology-enabled interventions which incorporatetraditional and empirically proven behaviour changemethodologies (as described in this work-in-progress),could eliminate the need for consistent and permanentfeedback to be delivered if the target behaviour hasbeen adopted.

This was a small scale pilot study and only a largerinvestigation could conclusively determine how effectiveaversive feedback may be when used in the context ofchanging behaviour in energy consumption. Building onour findings, we plan to run a longitudinal energy studydeploying aversive feedback with a view to evaluatingpotential behaviour change in energy consumption. Theevaluation would test whether aversive stimuli, chartmusic and exposing undesirable consumption in thiscase, could actually function as a punisher to motivatereductions in energy usage as a result of the feedback.

References[1] Ai He, H., Greenberg, S., and Huang, E.M. (2010) Onesize does not fit all: applying the transtheoretical model toenergy feedback technology design. Proc. CHI '10, 927-936[2] Bang, M., Torstensson, C., and Katzeff, C. (2006) ThePowerHouse: A Persuasive Computer Game Designed to RaiseAwareness of Domestic Energy Consumption, Lecture Notes inComputer Science, Vol. 3962, 123--132.

[3] Campbell, T., Ngo, B., and Fogarty, J. (2008) GameDesign Principles in Everyday Fitness Applications. Proc. Confon Computer Supported Cooperative Work,249-252.[4] Connected Environments Ltd. (2010) Pachube APIDocumentation, available from http://api.pachube.com/[5] Consolvo, S., Everitt, K., Smith, I., and Landay, J.A. (2006)Design Requirements for Technologies that EncouragePhysical Activity. Proc. CHI’06, 457-66.[6] Consolvo, S., McDonald, D.W., and Landay, J.A. (2009)Theory-driven design strategies for technologies that supportbehavior change in everyday life. Proc. CHI '09, 405-414.[7] Current Cost (2008) Current Cost – Reducing your energybills, available from http://www.currentcost.com/product-bridge.html[8] Darby, S. (2006) The effectiveness of feedback on energyconsumption. University of Oxford, Environmental ChangeInstitute[9] Facebook (2011) Facebook Graph API, available fromhttp://developers.facebook.com/docs/[10] Fogg, B.J. (2003) Persuasive Technology – UsingComputers to Change What We Think and Do. MorganKaufmann, San Francisco.[11] Froehlich, J., Consolvo, S., Dillahunt, T., Harrison, B.,Klasnja, P., Mankoff, J., Landay, J. (2009) Ubigreen:investigating a mobile tool for tracking and supporting greentransportation habits. Proc. CHI’09, 1043-1052.[12] Froehlich, J., Findlater, L., and Landay, J. (2010) Thedesign of eco-feedback technology. Proc. CHI’10, 1999-2008.[13] Kazdin, A. E. (2008). Behavior modification in appliedsettings (6th ed.). Belmont, CA: Wadsworth/ ThompsonLearning.[14] Kirman, B., Linehan, C., Lawson, S., Foster, D., andDoughty, M. (2010) There's a monster in my kitchen: usingaversive feedback to motivate behaviour change. Proc. CHI’10,2685-2694.[15] Yun, T. (2009) Investigating the impact of a minimalist in-home energy consumption display. Proc. CHI’09, 4417-4422.