=Paper=
{{Paper
|id=Vol-2450/short3
|storemode=property
|title=The Effectiveness of Advice Solicitation and Social Peers in an Energy Recommender System
|pdfUrl=https://ceur-ws.org/Vol-2450/short3.pdf
|volume=Vol-2450
|authors=Alain Starke
|dblpUrl=https://dblp.org/rec/conf/recsys/Starke19
}}
==The Effectiveness of Advice Solicitation and Social Peers in an Energy Recommender System==
https://ceur-ws.org/Vol-2450/short3.pdf
The effectiveness of advice solicitation and social peers in an
energy recommender system
Alain Starke
Human-Technology Interaction Group
Eindhoven University of Technology
Eindhoven, the Netherlands
A.D.Starke@tue.nl
ABSTRACT own behavior: “this is what I would do”, even though this might
not be as persuasive. In addition, whether one’s advice is
In face-to-face interactions, advice acceptance depends on how it accepted might actually depend on who is giving advice: a good
is presented, as well as a number of social factors. For example, friend or a stranger [2, 35].
some persons are inclined to accept advice from an expert if they Research has examined the effects of advice form, task
possess little domain knowledge. In contrast, if such advice is difficulty, and advice source on whether this advice is accepted
unsolicited, persons might only accept advice from a trusted by others [3, 6, 10]. Although each of these factors are important
source, such as a family member. Whether these mechanisms on their own [6], their interplay might affect advice acceptance
also play a role in the recommender context is unknown, even differently. For example, advice given by those who are similar
though advice solicitation may be particularly important in or who are trusted, is more likely to be accepted than advice
domains where a recommender user seeks behavioral change given by strangers [14]. However, if the task or decision at hand
(e.g. energy conservation, healthy eating). This study examines is difficult, one might be more likely to trust and to accept expert
the role of advice solicitation (i.e. whether one asks for advice or advice instead [9]. Nonetheless, this also depends on whether the
simply receives it) and advice source (i.e. either explained in advice is asked for or unsolicited [12, 13].
social terms or not) in our ‘Saving Aid’ energy recommender Such mechanisms of advice-taking are given little attention in
system. Through a web-based user study with 252 participants, an HCI context, but may impact whether a user actually acts on
we find that allowing users to solicit advice themselves increases presented recommendations. Most recommender systems tend to
their perceived level of trust with our energy recommender be prescriptive in presenting their content (“chosen for you”),
system, compared to users that are presented unsolicited advice. and do not consider whether some users prefer to browse for
In turn, we find that trust positively affects user satisfaction appropriate content instead. Moreover, although it has been
levels, as well as the number of chosen energy-saving measures. shown that explaining recommendation in terms of social peers
We discuss how system designers should consider how advice is increases the likelihood that items are chosen [4], as well as
presented and in which context. evaluated satisfactorily [17, 28], these findings are usually
limited to social network applications.
CCS CONCEPTS The interplay between advice solicitation and advice source
• Informa on Systems Decision Support Systems might be particularly important in domains where behavioral
• Human‐centered Compu ng User Studies change is part of the recommender ecosystem, such as energy
conservation and health [18, 25, 29, 33]. There, personalization is
KEYWORDS merely the starting point of persuading a user to take up a new
habit [7, 24, 31]. For example, besides predicting which specific
Recommender Systems, Behavioral Change, Energy energy-saving measures are appropriate for a user, the use of
Conserva on, User Experience, Advice Solicita on social comparison and peer feedback might ultimately persuade
that user to change his or her energy conservation habits [23,
1 Introduction 27]. Moreover, there is evidence that users might only be willing
When one gives another person advice, it is often worded in no to change their habits if they have solicited advice themselves,
uncertain terms what the other person should do, for example by instead of the advice being ‘forced’ upon them [12].
saying “You should buy this item”. Such advice may be Furthermore, energy recommender systems need to consider
unsolicited by the other person and, therefore, not accepted. In that each behavior has a different execution difficulty [25, 29].
other situations, much more precaution is taken by sharing one’s An effective energy recommender should persuade its users to
choose and perform more energy-efficient behaviors [18, 29, 32].
IntRS ’19: Joint Workshop on Interfaces and Human Decision Making for Whether a user is willing to only make a small change in his
Recommender Systems, September 19, 2019, Copenhagen, Denmark. conservation habits (e.g. changing one’s light bulbs), or a large
Copyright © 2019 for this paper by its authors. Use permitted under Creative
one (e.g. installing solar PV), may depend on which other users
Commons License Attribution 4.0 International (CC BY 4.0).
have already adopted a certain behavior [1, 23, 27].
IntRS ’19, September 19, 2019, Copenhagen, Denmark A.D. Starke
1.1 Research Question Furthermore, unsolicited advice also tends to be accepted
when it is given by a friend or family member [10, 12]. However,
This paper examines whether advice solicitation affects how a
some users might feel as if the added social sources are ‘butting
recommender system and its advice are evaluated. For example,
in’, an experience that is often associated with unsolicited aid
some users benefit from browsing a recommender’s personalized
and which usually leads to poor rates of advice acceptance [3, 8].
interface [15], without requiring an explicit description of each
Advice solicitation impacts advice acceptance, regardless of
item’s fit. Moreover, whether users wish to solicit advice
whether the advice is of high quality or not. According to Fisher
themselves or that unsolicited advice is accepted, may depend on
et al. [8], negative responses to aid (e.g. recommended items)
the advice’s source. In this case, we examine whether the advice
often stem from the recipient’s perceived threat to self-esteem or
is explained in social terms or not, coming from either the
autonomy. This is conceptualized as ‘threat to face’ [12]: positive
system itself, a similar user, or an expert.
face equals one’s positive self-image, whereas negative face
We posit the following research question:
relates to one’s autonomy and control over one’s own life [8].
[RQ]: To what extent do advice solicitation and social advice
Threats to both types of face are lower when an advice recipient
affect a user’s perception and evaluation of an energy
(or user) seems to have asked for advice [12]. Threat to face
recommender system?
seems to be more important in conflict-avoiding cultures, such as
In the upcoming section, we discuss which psychological
those found in eastern Asia [21].
concepts underlie advice-giving and taking, and how advice
In the context of an energy recommender system, the
solicitation and social advice might influence these. In addition,
addition of social sources might only work if the user has an
to generate energy recommendations, we present a ‘light
option to solicit advice or not, as it might otherwise pose a threat
personalization’ algorithm using the psychometric Rasch model.
to one’s autonomy and, in turn, decrease the trust and quality of
In section 3, we present our ‘Saving Aid’ recommender system.
the recommended advice. Hence, we expect that allowing users
to solicit advice leads to different evaluations of a recommender
2 Related Work system in terms of trust and quality, compared to a system that
presents unsolicited advice. Moreover, we expect this effect to
2.1 Advice form, acceptance, and autonomy
depend on how the advice is explained, using either social peers
When someone receives advice in a face-to-face situation, three or not.
different motives on the advisee’s part are found to be important
[3, 26]: increasing one’s decision accuracy, minimizing decision 2.2 Personalization in energy recommenders
effort, and maintaining autonomy. Whereas the first two are
Our research question is contextualized in the household energy
typically addressed by recommender systems in the human-
domain. Previous energy recommender studies have shown that
computer interaction domain, the latter has received less
a simple personalization algorithm, based on the psychometric
attention in a personalization context.
Rasch model, can lead to positive changes in user evaluations
One’s autonomy in decision-making has traditionally been
[29]. This item response theory model assumes that all energy-
defined as one’s need to resist influence or coercion, or to strive
saving measures and persons share a one-dimensional trait for
for independence [22]. Although this begs the question why
the goal of saving energy [34], which manifests itself as a single
persons with a high need for autonomy would seek advice from
measurement scale.
others [20, 22], let alone recommender systems, much of the
The Rasch model is an operationalization of ‘Campbell’s
advice and personalization that one faces in life is unsolicited
Paradigm’ [16]. This attitude theory postulates that one’s
[12]. Even so, it is arguably inevitable that also a highly-
energy-saving attitude becomes apparent through the behavioral
autonomous person requires advice from others for complex
steps that one is willing to take. Put differently, the more
decisions where he or she lacks domain knowledge.
measures one takes with the goal of saving energy, the stronger
How advice is evaluated, appears to depend on the interplay
one’s energy-saving attitude is expected to be [16, 29, 34]. In a
between how advice representation and the advisee’s autonomy
similar vein, these energy-saving measures are assumed to differ
[5, 6]. For instance, prescriptive advice (‘you should do X’) is
in how difficult they are to perform, which is operationalized as
typically evaluated worse than descriptive advice (‘I would do
their behavioral costs [16]. Measures that are performed less
X’), if the person has a high need for autonomy [6]. However, if
often are assumed to have higher behavioral costs, and vice
the advice source has a high level of perceived expertise,
versa [16, 34].
unsolicited and prescriptive advice tends to be evaluated more
The Rasch model is described in Equation 1. A person n with
favorably, compared to advice coming from a stranger [6, 13].
an attitude θn that exceeds the behavioral costs δi of an energy-
This may be particularly important when a user lacks the
saving measure i, has a high probability P of performing such
capabilities to make an accurate decision [20], causing the user’s
measure: P {Xni = 1}. In contrast, that probability is expected to be
need for autonomy to diminish. For example, a recommender
low if that person’s attitude is much lower than the behavioral
system that presents items that are unfamiliar to a user, might
costs of the measure at hand:
want to explain its recommendations in terms of other expert
users. 1 (1)
Advice solicitation and social peers in energy recommender systems IntRS ’19, September 19, 2019, Copenhagen, Denmark
Using the Rasch model, Starke et al. [29] have reliably fitted a savings (NL: ‘Besparing’) of each measure were listed. To
one-dimensional scale of 79 energy-saving measures. These are personalize the Saving Aid, it first presented the best fitting
either one-time investments (“insulate the exterior walls”), or measures according to the Rasch model (cf. Equation 1),
frequent curtailment behaviors (“turn off lights after leaving a presenting those that were closest to a user’s attitude in terms of
room”). The construct is used in the current study to present a their behavioral costs. Users were then asked to choose (NL:
list of personalized energy-saving measures that strike the right ‘Kies’) any number of measures they would like to perform in
balance between attractiveness (the behavioral costs are not too the weeks following the study. As soon as users had finished
high for the user’s attitude) and novelty (the user is expected to navigating the interface and had chosen measures, they were
not already perform each measure), by minimizing the difference presented a questionnaire on how they perceived the presented
between a user’s energy-saving attitude and a measure’s recommendations, as well as how they evaluated the system.
behavioral costs. If attitude and behavioral costs are equal, a
measure’s engagement probability is 50% (cf. Equation 1).
3 Method
To investigate how advice solicitation and social advice
explanations affect a user’s evaluation of an energy
recommender system, we performed a between-subjects web
study. To do so, we developed the ‘Saving Aid’ recommender
system (cf. Figure 2), a free-to-use ‘web shop’ that presented
attitude-tailored energy-saving measures based on a user’s past
behavior and the Rasch model. Users could choose any number
of measures they wished to perform in the weeks following the
study, which would be sent to them by email.
3.1 Procedure
Figure 1 depicts the general procedure of the current
recommender study, which was similar to the procedure used in
an earlier study of Starke et al. [29]. First, we estimated each
user’s energy-saving attitude, by surveying their current energy-
saving behavior. Similar to other studies [30, 31], we used a short
survey of thirteen energy-saving measures. To do so, we divided
the Rasch scale in thirteen subsets across its entire behavioral Figure 2. Interface of the ‘Saving Aid’ energy
cost range (from δ = −4.41 to δ = 4.42; Mδ = 0.05), and randomly recommender system (in Dutch). 79 measures were
sampled a measure from each subset. For each measure, users presented in ascending order of behavioral costs (labeled
had to indicate whether they already performed it (‘yes’ or ‘no’), as ranging from ‘easy/popular’ on the left, to
or that a measure did not apply to their housing situation (e.g. ‘challenging/less popular’ on the right), and could be
energy-efficient garden lighting did not apply to a user who did freely navigated using the colored blocks. Each measure
listed its name (e.g. “Install double-glazed windows” on the
not own a garden).
left), as well as its investment costs (NL: “Kosten”) and
The estimated attitude θ was based on the number of ‘yes’
projected kWh Savings (NL: “Besparing”). The text at the
responses. We used the average behavioral cost level δ of the top: “klik hier […] u aanraden,” translates to “click here if
equivalent Rasch scale subset. For example, if a user had you would like to see what others recommend you”.
submitted six ‘yes’ responses, the attitude was estimated to be
equal to the average behavioral costs of the sixth subset. 3.2 Research Design
How the advice and the recommender interface were presented,
was subject to a 3x2 between-subjects research design. For each
condition, we changed how advice was presented. On the one
hand, we discerned between three different sources of advice:
pointing out that specific measures were recommended by either
Figure 1. Procedure of the current recommender study. an expert, similar users, or the system itself. These would be
added as explanations to the system’s solicitation button (cf. top
Subsequently, users were navigated to the ‘Saving Aid’ of Figure 2) or the unsolicited measures (cf. Figure 3).
interface, which is shown in Figure 2. It presented the Rasch On the other hand, we differentiated between two levels of
scale of 79 energy-saving measures in ascending order of advice solicitation. In the unsolicited advice condition, users
behavioral costs (from ‘popular’ to ‘challenging’), which users were immediately presented three different energy-saving
were free to navigate. The name, costs (NL: ‘Kosten’) and kWh measures in the pop-up screen depicted in Figure 3. They were
IntRS ’19, September 19, 2019, Copenhagen, Denmark A.D. Starke
free to choose any of these measures, or could continue to the 3.4 Measures
main interface. In the solicited advice condition, users were only
shown a pop-up that explained how the interface worked, after 3.4.1. Objective aspects
which they could continue to the main Saving Aid interface. The different recommender conditions (presenting either
However, users could solicit personalized recommendations in solicited or unsolicited advice, from either a similar source, an
the main interface by clicking a button, which is depicted at the expert, or the system itself) were considered as objective changes
top of Figure 2: ‘click here if you want to see what others would in the interface. We tested whether these changes affected how
recommend to you’. users perceived the presented recommendations and, in turn,
evaluated the system. Furthermore, we also examined whether
this led to changes in the number of chosen measures.
Upon analysis, we observed that only half of all users in the
‘solicited’ condition had actually clicked to solicit personalized
recommendations (cf. Figure 2). Since we expected that either
inspecting a recommendation list or not could affect how users
perceived and evaluated the system, we decided to discern
between three groups of solicitation instead: unsolicited advice
(i.e. baseline), no advice solicitation (did not click), and advice
solicitation. Since testing a 3x3 design with a sample of 252
participants would only allow us to detect rather large effects,
we collapsed the three social conditions (system, similar, or
expert) into two: advice from either a social or non-social source.
This allowed us to interpret the results with sufficient statistical
power. Note that we found no significant differences between
similar and expert advice in a separate analysis.
Figure 3. The pop-up screen in the unsolicited condition, 3.4.2. Subjective aspects
which presented energy-saving advice immediately after We surveyed users on five subjective constructs: perceived
attitude calibration. Similar to Figure 2, each measure’s recommendation quality, perceived system trust, threat to
name, costs, and kWh savings was listed. This example negative face (i.e. whether one’s autonomy is affected), system
depicts advice explained through an expert source (NL:
satisfaction, and choice satisfaction. For all these aspects, users
“expert advies”). Cropped off the top of the image is an
explanatory text in Dutch. Users could choose (NL: “Kies”) were presented survey items on 7-point Likert scales and were
any of these measures if they wished to perform them. asked to indicate to what extent they agreed with each item. The
used questionnaire items are described in Table 1.
As prescribed by Knijnenburg and Willemsen [19], we
3.3 Participants
submitted all responses to a confirmatory factor analysis (CFA)
In total, 260 participants used our Saving Aid recommender using ordinal dependent variables. Table 1 reports only three
system and finished the subsequent questionnaire. Among them, user experience constructs, as we could not include the threat to
110 participants were recruited from the Jan Frederik Schouten face construct in our analysis, for it had high cross loadings with
database of Eindhoven University of Technology, while others all other aspects in our model. Moreover, we could not discern
were recruited through posts on social media (e.g. Twitter, choice satisfaction from the system satisfaction construct, since
Facebook, etc.) Each participant entered a raffle in which they it violated divergent validity [19]. The remaining constructs met
had a 20% probability to win €15. We omitted 8 participants from the guidelines for convergence validity, as the average variance
analysis, as they either had finished the study in no more than 2 explained (AVE) for each construct was higher than 0.5, and had
minutes, or showed no variation in their answers on the a good internal consistency (0.8 < α < 0.9) [11].
evaluation questionnaire.
Eventually, we analyzed a sample of 252 participants (50.8%
4 Results
female), which had a mean age of 31.9 years (SD = 16.0), and who
on average chose 8.9 measures (SD = 9.1). Furthermore, We used Structural Equation Modeling (SEM) to organize all
approximately half of the sample was still a university student, objective and subjective constructs, including relevant
who typically lived in shared apartments. Most of these students interactions, into a path model. As prescribed by Knijnenburg
had an income that fell below €1000. and Willemsen [19], a confirmatory factor analysis was
performed first (cf. Table 1), after which we tested a fully
saturated model and performed stepwise removal of non-
significant relations. Figure 4 depicts the final path model, which
had an excellent fit: χ2(113) = 131.625, p = 0.11, CFI = 0.995, TLI =
0.996, RMSEA = 0.026, 90%-CI: [0.000,0.042].
Advice solicitation and social peers in energy recommender systems IntRS ’19, September 19, 2019, Copenhagen, Denmark
Table 1. Results of the confirmatory factory analysis on satisfaction showed that this effect was significant, mediated by
user experience. Items without loading were removed perceived trust: β = .151, 95%-CI: [0.004,0.298], p = .044.
from the final model, while the Choice Satisfaction and Second, Figure 4 also shows two different effects of users who
Threat to Negative Face aspects were excluded as they were in the solicited condition, but who did not solicit advice. If
violated divergent validity. The average variance such advice was offered using a socially-laden explanation
explained (AVE) and Cronbach’s Alpha of other aspects
(either through expert advice or similar peers), it negatively
met the prescribed guidelines [11, 19].
impacted the perceived recommendation quality compared to the
unsolicited, non-social baseline: β = −1.18, p < 0.01. In contrast,
Aspect Item Loa‐ non-social explanations positively affected the perceived
ding
recommendation quality compared to the unsolicited baseline: β
Choice I am happy with the measures I’ve chosen
Satisfaction I know several measures that are better than = .41, p < 0.05. Since non-soliciting users did not actually see a
the ones I selected list of highlighted energy-saving recommendations, it was
I would recommend some of the chosen possible they evaluated the main interface instead, which
measures to others confounds a clear interpretation of this result.
I am looking forward to implement the chosen
measures
The measures I’ve chosen fit me seamlessly
Perceived I found the recommended measures to be .813
Rec. Quality attractive
The recommended measures fitted my .857
AVE: .68 preferences
Alpha: .82 The recommended measures were relevant to .731
me
The Saving Aid proposed too many bad
measures
I did not like any of the measures
Perceived I think that the Saving Aid was telling me the .756
Sys. Trust truth
I expected the Saving Aid to be truthful .670
AVE: .72 The Saving Aid was honest .747
Alpha: .88 The Saving Aid was sufficiently Figure 4. Structural Equation Model (SEM). The numbers
knowledgeable to present advice on the arrows represent β-coefficients, standard errors are
The Saving Aid had the best intentions .624 denoted between brackets. The effects between the latent
subjective constructs are standardized and can be
System The Saving Aid has made me more aware of .546 considered as correlations. Aspects are grouped by color:
satisfaction my energy-saving behavior objective system aspects are purple, interaction aspects are
I would like to use the Saving Aid more often .663 blue, subjective aspects are green, and experience aspects
AVE: .70 I make better decisions using the Saving Aid .710
are orange. *** p < 0.001, ** p < 0.01, * p < 0.05.
Alpha: .88 The Saving Aid helps me to find appropriate .620
measures
The Saving Aid allowed me to choose 4.2 System evaluation and choice behavior
measures easily
Figure 4 depicts that users who had chosen a measure from
Threat to The Saving Aid respected my autonomy and either a solicited or unsolicited recommendation list, perceived
Negative the choices I made the recommendation quality to be higher than those who did
Face The Saving Aid did not impose anything on not: β = .636, p < 0.001. Furthermore, Figure 4 shows two
me
I was free to choose any measure different pathways in which higher levels of recommendation
I did not feel forced to take the Saving Aid’s quality increase other subjective aspects. First, higher perceived
advice recommendation quality positively affected perceived trust and,
in turn, system satisfaction. Besides this mediated effect, Figure 4
4.1 Advice solicitation also shows a positive, direct path from recommendation quality
Figure 4 shows two effects of advice solicitation on how the to system satisfaction, as well as a positive effect from users
recommender system and its advice were perceived. First, users choosing a recommended measure to system satisfaction.
who had the option to solicit advice and actually did so (‘advice Moreover, a positive evaluation of the system also led users to
solicitation’), reported higher levels of system trust than those choose more energy-saving measures: β = .979, p < 0.05.
who were presented unsolicited advice: β = .414, p < 0.05. This Although it could be possible to reverse this particular causal
suggested that users who asked for personalized advice direction (e.g. more choices led to higher satisfaction levels), the
perceived the system as more trustworthy, than those who current pathway was consistent with previous energy
immediately faced unsolicited advice, regardless of the social recommender studies [18, 29], and led to the best model fit
source. A bootstrapped test of indirect effects towards system statistics.
IntRS ’19, September 19, 2019, Copenhagen, Denmark A.D. Starke
5 Discussion 5.1 Limitations
As one of the first to do so, this study has applied the Unfortunately, we could not relate our path model constructs to
phenomenon of face-to-face advice solicitation (asking for advice choice satisfaction, for it violated divergent validity. However,
or not) to the HCI domain. We have investigated to what extent the strong correlation between both system and choice
solicitation and social explanations of advice affect how a user satisfaction suggests that trust would also have a positive effect
perceives and evaluates an energy recommender system. on how users evaluate their choices, which has shown to
To do so, we have developed the ‘Saving Aid’ system, from positively affect the probability that chosen measures are
which users could choose any number of energy-saving actually implemented and a user’s behavior is changed [29].
measures they would like to perform. It employs a simple Our findings are confounded by not all users soliciting advice
personalization algorithm, using the psychometric Rasch model, in the ‘solicitation condition’. This has prevented us from
to generate appropriate attitude-tailored recommendations for analyzing our original research design, which has now
each user. In this personalized advice context, we have found overlooked any differences between specific social explanations,
small differences in how our energy recommender system is either a similar peer or an expert. Although this should be
evaluated, based on whether advice is solicited by users or further investigated in a follow-up study, the discrepancy
presented without being explicitly solicited. between social and non-social explanations of advice seems to be
Our results show that users who have solicited personalized an important finding in itself, which is also demonstrated by the
recommendations report higher system trust levels, compared to interaction effect in our path model.
those who are presented unsolicited advice. It could be that users Furthermore, in studies on face-to-face interactions, advice
wish to maintain control over which items they inspect, and that solicitation is typically related to personality concepts as
a system that immediately determines which three items are autonomy and threat to face. Due to a violation of divergent
appropriate, is perceived as less trustworthy. The fact that most validity in our path model, we have been unable to test whether
recommender research does not consider whether some users a violation of autonomy has decreased trust. Nonetheless, our
wish to solicit advice themselves, seems to be a missed study still shows that changes in the recommender interface, due
opportunity, since higher trust levels in a recommender could to advice solicitation and social explanations, lead to changes in
have important second-order effects. Indeed, in this study, we system perception and choice behavior, and should be
find that users who report higher trust levels, also choose more considered in future recommender designs.
energy-saving measures as a result of a positive user experience.
Furthermore, we find mixed results for perceived ACKNOWLEDGMENTS
recommendation quality across social conditions. Users who We thank Nena van As, Jochem Bek, and Boy Janissen for their
have the opportunity to solicit advice but do not do so, report time and effort to conduct this study and discuss its implications.
lower levels of recommendation quality if advice is explained in This work is part of the Research Talent program with project
terms of similar peers or experts, compared to users who face number 406-14-088, which is financed by the Netherlands
unsolicited social advice. This effect reverses for non-social Organization for Scientific Research (NWO). It was also made
explanations, as users who are presented the opportunity to possible by funding from the Niels Stensen Fellowship.
solicit system advice report higher recommendation quality than
those who are presented unsolicited system advice. These REFERENCES
findings either suggest that adding social explanations to [1] Allcott, H. 2011. Social norms and energy conservation. Journal of Public
unsolicited advice could mitigate a user’s feeling that the system Economics. 95, 9–10 (Oct. 2011), 1082–1095.
is ‘butting in’, or that users who could solicit advice might not be DOI:https://doi.org/10.1016/j.jpubeco.2011.03.003.
[2] Bo Feng and MacGeorge, E.L. 2010. The Influences of Message and Source
interested in additional social advice after inspecting the main Factors on Advice Outcomes. Communication Research. 37, 4 (Aug. 2010),
Saving Aid interface. However, these comparisons only apply to 553–575. DOI:https://doi.org/10.1177/0093650210368258.
[3] Bonaccio, S. and Dalal, R.S. 2006. Advice taking and decision-making: An
those who have not clicked, which has probably increased the integrative literature review, and implications for the organizational
differences between conditions. sciences. Organizational Behavior and Human Decision Processes. 101, 2
(Nov. 2006), 127–151. DOI:https://doi.org/10.1016/j.obhdp.2006.07.001.
Our results resonate with earlier findings on inspectability [4] Bonhard, P. and Sasse, M.A. 2006. ’Knowing me, knowing you’ — Using
and control in social recommender systems [17], which has profiles and social networking to improve recommender systems. BT
shown to positively affect system evaluation. Furthermore, it Technology Journal. 24, 3 (Jul. 2006), 84–98.
DOI:https://doi.org/10.1007/s10550-006-0080-3.
shows that findings from the advice-giving and taking literature [5] Caplan, S.E. and Samter, W. 1999. The role of facework in younger and
on face-to-face interactions do translate to the HCI context [e.g. older adults’ evaluations of social support messages. Communication
Quarterly. 47, 3 (1999), 245–264.
3, 12]. Even though the interactions in a typical recommender [6] Dalal, R.S. and Bonaccio, S. 2010. What types of advice do decision-makers
system are hardly anthropomorphic, the principles of trust, prefer? Organizational Behavior and Human Decision Processes. 112, 1 (May
quality, and possibly autonomy might also translate to 2010), 11–23. DOI:https://doi.org/10.1016/j.obhdp.2009.11.007.
[7] Ekstrand, M.D. and Willemsen, M.C. 2016. Behaviorism is Not Enough:
personalized HCI. These factors may be particularly important in Better Recommendations Through Listening to Users. Proceedings of the
domains where users wish to change their lifestyle, such as 10th ACM Conference on Recommender Systems (New York, NY, USA, 2016),
221–224.
health and energy conservation, which often rely on social proof [8] Fisher, J.D., Nadler, A. and Whitcher-Alagna, S. 1982. Recipient reactions
to achieve behavioral change [1, 27]. to aid. Psychological Bulletin. 91, 1 (1982), 27.
Advice solicitation and social peers in energy recommender systems IntRS ’19, September 19, 2019, Copenhagen, Denmark
[9] Gino, F. and Moore, D.A. 2007. Effects of task difficulty on use of advice. [33] Trattner, C. and Elsweiler, D. 2017. Food Recommender Systems:
Journal of Behavioral Decision Making. 20, 1 (Jan. 2007), 21–35. Important Contributions, Challenges and Future Research Directions.
DOI:https://doi.org/10.1002/bdm.539. arXiv:1711.02760 [cs]. (Nov. 2017).
[10] Gino, F., Shang, J. and Croson, R. 2009. The impact of information from [34] Urban, J. and Ščasný, M. 2016. Structure of Domestic Energy Saving: How
similar or different advisors on judgment. Organizational Behavior and Many Dimensions? Environment and Behavior. 48, 3 (Apr. 2016), 454–481.
Human Decision Processes. 108, 2 (Mar. 2009), 287–302. DOI:https://doi.org/10.1177/0013916514547081.
DOI:https://doi.org/10.1016/j.obhdp.2008.08.002. [35] Yaniv, I., Choshen-Hillel, S. and Milyavsky, M. 2011. Receiving advice on
[11] Gliem, J.A. and Gliem, R.R. 2003. Calculating, Interpreting, And Reporting matters of taste: Similarity, majority influence, and taste discrimination.
Cronbach’s Alpha Reliability Coefficient For Likert-Type Scales. (2003). Organizational Behavior and Human Decision Processes. 115, 1 (May 2011),
[12] Goldsmith, D.J. 2000. Soliciting advice: The role of sequential placement in 111–120. DOI:https://doi.org/10.1016/j.obhdp.2010.11.006.
mitigating face threat. Communication Monographs. 67, 1 (Mar. 2000), 1–19.
DOI:https://doi.org/10.1080/03637750009376492.
[13] Goldsmith, D.J. and Fitch, K. 1997. The Normative Context of Advice as
Social Support. Human Communication Research. 23, 4 (Jun. 1997), 454–476.
DOI:https://doi.org/10.1111/j.1468-2958.1997.tb00406.x.
[14] Harvey, N. and Fischer, I. 1997. Taking Advice: Accepting Help, Improving
Judgment, and Sharing Responsibility. Organizational Behavior and Human
Decision Processes. 70, 2 (May 1997), 117–133.
DOI:https://doi.org/10.1006/obhd.1997.2697.
[15] Herlocker, J.L., Konstan, J.A., Terveen, L.G. and Riedl, J.T. 2004. Evaluating
Collaborative Filtering Recommender Systems. ACM Trans. Inf. Syst. 22, 1
(Jan. 2004), 5–53. DOI:https://doi.org/10.1145/963770.963772.
[16] Kaiser, F.G., Byrka, K. and Hartig, T. 2010. Reviving Campbell’s Paradigm
for Attitude Research. Personality and Social Psychology Review. 14, 4 (Nov.
2010), 351–367. DOI:https://doi.org/10.1177/1088868310366452.
[17] Knijnenburg, B.P., Bostandjiev, S., O’Donovan, J. and Kobsa, A. 2012.
Inspectability and control in social recommenders. Proceedings of the sixth
ACM conference on Recommender systems (2012), 43–50.
[18] Knijnenburg, B.P., Willemsen, M. and Broeders, R. 2014. Smart
sustainability through system satisfaction: Tailored preference elicitation
for energy-saving recommenders. Proceedings 20th Americas Conference on
Information Systems (AMCIS 2014): Smart Sustainability: The Information
Systems Opportunity (Savannah, Georgia, United States, 2014).
[19] Knijnenburg, B.P. and Willemsen, M.C. 2015. Evaluating recommender
systems with user experiments. Recommender Systems Handbook. Springer.
309–352.
[20] Koestner, R., Gingras, I., Abutaa, R., Losier, G.F., DiDio, L. and Gagné, M.
1999. To Follow Expert Advice When Making a Decision: An Examination
of Reactive Versus Reflective Autonomy. Journal of Personality. 67, 5
(1999), 851–872. DOI:https://doi.org/10.1111/1467-6494.00075.
[21] Meyer, E. 2014. The culture map: Breaking through the invisible boundaries
of global business. Public Affairs.
[22] Murray, H.A. 1938. Explorations in personality: A clinical and
experimental study of fifty men of college age. (1938).
[23] Nolan, J.M., Schultz, P.W., Cialdini, R.B., Goldstein, N.J. and Griskevicius,
V. 2008. Normative social influence is underdetected. Personality and social
psychology bulletin. 34, 7 (2008), 913–923.
[24] Schäfer, H., Hors-Fraile, S., Karumur, R.P., Calero Valdez, A., Said, A.,
Torkamaan, H., Ulmer, T. and Trattner, C. 2017. Towards Health (Aware)
Recommender Systems. Proceedings of the 2017 International Conference on
Digital Health (New York, NY, USA, 2017), 157–161.
[25] Schäfer, H. and Willemsen, M.C. 2019. Rasch-based Tailored Goals for
Nutrition Assistance Systems. Proceedings of the 24th International
Conference on Intelligent User Interfaces (New York, NY, USA, 2019), 18–29.
[26] Schrah, G.E., Dalal, R.S. and Sniezek, J.A. 2006. No decision-maker is an
Island: integrating expert advice with information acquisition. Journal of
Behavioral Decision Making. 19, 1 (Jan. 2006), 43–60.
DOI:https://doi.org/10.1002/bdm.514.
[27] Schultz, P.W., Nolan, J.M., Cialdini, R.B., Goldstein, N.J. and Griskevicius,
V. 2007. The constructive, destructive, and reconstructive power of social
norms. Psychological science. 18, 5 (2007), 429–434.
[28] Sharma, A. and Cosley, D. 2013. Do social explanations work?: studying
and modeling the effects of social explanations in recommender systems.
Proceedings of the 22nd international conference on World Wide Web (2013),
1133–1144.
[29] Starke, A., Willemsen, M. and Snijders, C. 2017. Effective User Interface
Designs to Increase Energy-efficient Behavior in a Rasch-based Energy
Recommender System. Proceedings of the Eleventh ACM Conference on
Recommender Systems (New York, NY, USA, 2017), 65–73.
[30] Starke, A., Willemsen, M.C. and Snijders, C. 2015. Saving Energy in 1-D:
Tailoring Energy-saving Advice Using a Rasch-based Energy
Recommender System. DMRS (2015), 5–8.
[31] Starke, A.D. 2019. Supporting energy-efficient choices using Rasch-based
recommender interfaces. (2019).
[32] Tomkins, S., Isley, S. and Getoor, L. 2018. Sustainability at Scale: Towards
Bridging the Intention-Behavior Gap with Sustainable Recommendations.
Proceedings of the 12th ACM Conference on Recommender Systems
(Vancouver, British Columbia, Canada, 2018), 214–218.