Effect of Simulation and Social Comparison Feedback in
Food Waste Reduction Applications
SUBMITTED IN PARTIAL FULLFILLMENT FOR THE DEGREE OF MASTER
OF SCIENCE
A
NETA
P
ENEVA
10672257
M
ASTER
I
NFORMATION
S
TUDIES
H
UMAN-
C
ENTERED
M
ULTIMEDIA
F
ACULTY OF
S
CIENCE
U
NIVERSITY OF
A
MSTERDAM
A
UGUST
21, 2015
1st Supervisor 2nd Supervisor
Dr. Frank Nack Dr. Lynda Hardman
Effect of Simulation and Social Comparison Feedback in
Food Waste Reduction Applications
Aneta Peneva (10672257)
ABSTRACT
In developing countries, end consumers are the greatest con-tributors to the food waste problem. Based on the premise that the practical reasons for this behaviour seem to be eas-ily avoidable, we assumed that if consumers have the inten-tion to change, they would do so. We designed and tested two interfaces, aiming at influencing their intentions. Both provided information about food waste’s environmental im-pact but enhanced in different ways. One used simulation to show accumulation of food waste’s impact over time. The other used social comparison - displayed information about the intentions for food waste reduction of others. We hy-pothesized that simulation and high social norms would in-crease intentions and that a low norm would lead to a de-crease. Our experiment showed only insignificant changes in all cases. However, the provided awareness information pro-voked our users to state high food waste reduction aims. Additionally, simulation was perceived better than social comparison feedback. When provided with a high social norm, users stated the highest food waste reduction aims and low norm provided the lowest results.
General Terms
Behaviour change
Keywords
Application design, simulation, social comparison, social norms, food waste
1.
INTRODUCTION
Globally, it was estimated that around 1/3 of the food pro-duced for human consumption is lost or wasted each year, equal to about 1.3 billion tons [20]. At the same time, 925 million people are chronically hungry [19].
Besides being a potential contributor to the world hunger problem, food waste has serious impact on the environment. Food production and food waste disposal generate green-house gas emissions, mainly CO2, methane and nitrous ox-ide, which contribute to the global climate change (about 20% of the global CO2 emissions [7]. Additionally, food production involves enormous water consumption - it is es-timated that producing one calorie of food requires one liter of water [21].
Food waste’s economic implications are worth noting as well - UK citizens annually throw away food worth 12 billion euros every year, around 480 euros per household [5]; in the
US this ’luxury’ costs an average family of four between 1365 and 2275 dollars [19].
Food is lost and wasted all along the Food Supply Chain – from agricultural production to consumption [26]. Con-sumers generate a surprisingly large amount of this waste, especially in developed countries. “For affluent economies, post-consumer food waste accounts for the greatest overall losses” [26, p.3065]. The per capita food waste in Europe and North America amounts to 95 – 110 kg per year [20]. In the United States, consumers throw away about 25% of the food they buy; for the UK the percentage is the same [26]; for the Netherlands this is 8 – 11% [32].
According to [26], the greatest potential for food waste re-duction in developed countries lies within changing con-sumer behavior. Following this assumption, we investigate in this thesis ways to raise awareness of the problem of food waste to give people a start into changing their be-haviour [3, 15].
In this report, we first will discuss the reasons for food wasting behaviour as identified in the literature, the cur-rent technological solutions addressing them, and will look into behaviour change theory. We then outline the problem statement and define the research questions of the study. Afterwards, we explain the application prototypes designed and the experiment we conducted to evaluate them. We will also describe the collected data and the way it was pro-cessed. Finally, we discuss the results, the limitations of our approach, and opportunities for future work.
2.
RELATED WORK
2.1
Practical causes for consumer food waste
Lack of practical skills and basic knowledge related to food management have been associated with food wasting be-haviours. Lyndhurst [24] found out that people who re-ported poorer cooking skills waste larger amounts. Parfitt et al. [26] identify also the lack of knowledge about food stor-age, bad shopping habits (no shopping lists; impulse buy-ing); and failure to plan meal sizes among the reasons to waste food. Additionally, confusion around food labels (no distinction between “Best before” and “Use by” dates) and fear of food poisoning push people to throw away edible food [17, 29]. Finally, inconvenience is also a factor - people buy large amounts or cook more than they can consume with the initial purpose of saving time [12].
2.2
Lack of knowledge
For diverse reasons people do not prioritize food waste as a problem that needs to be addressed. Consumers appear to be unaware of the amount they waste - [24] discovered that 44% (N=1862) are not bothered by the issue as they consider they waste very little; [29] had similar findings (40%, N=2939). A reason for this attitude problem might be that it happens bit by bit [25].
Additionally, both studies reveal that many consumers (61% [29]; 39% [24]) are unaware of the environmental conse-quences related to food waste, as it seems natural and biodegrad-able. It is also notable, that even environmentally conscious respondents were only slightly less likely to agree that food has no impact on the environment (34% as compared to 39% of all participants in [24]). Gunders [19] adds “devaluation of food” to the lack of awareness list, arguing that we do not care if we throw away food, as in the last century our ability to buy more is only growing.
Current research is inconclusive on which information source would be more motivating – receiving information about cost of wasted food or about its environmental consequences. More of [29]’s respondents - 58.8% (N=2982) - agreed that information about environmental consequences would help them compared to 51.1% (N=2982), who stated this for in-formation about the financial consequences. According to several other studies, environmental concerns are less mo-tivating than saving money (20% aginst 68% in [24]; 24% against 74% in [27]). However, these authors also acknowl-edge the fact that users were unaware of food waste’s nega-tive impact on the environment.
In our review we retrieved only one article [13] - a small qualitative study (n=33), which gives an indication that, if provided with such information, consumers’ attitudes can shift positively. The participants received environmental information during 3-hour long discussions. After the dis-cussions, all 33 participants agreed that they are concerned about the environmental consequences of food waste, as op-posed to 16 before the focus groups.
2.3
Other food wasting issues
Tucker [33] found out that the more widespread and visible a pro-environmental action is, the more aware people be-come of their actions and more critical towards a diversion from these pro-environmental behaviours. Moreover, “The assurance that others are also doing their bit can be an im-portant motivator” [33, p. 11]. Food waste prevention is less “visible” than other pro-environmental behaviours (re-cycling, second hand shopping) and this is an obstacle to using social pressure to steer behaviour. Two qualitative studies [17, 27] confirm that people do not engage in food waste minimization because they are convinced no one else is doing it.
2.4
Current solutions
Tucker [33] found out that the more widespread and visible a pro-environmental action is, the more aware people be-come of their actions and more critical towards a diversion from these pro-environmental behaviours. Moreover, “The assurance that others are also doing their bit can be an im-portant motivator” [33, p. 11]. Food waste prevention is
less “visible” than other pro-environmental behaviours (re-cycling, second hand shopping) and this is an obstacle to using social pressure to steer behaviour. Two qualitative studies [17, 27] confirm that people do not engage in food waste minimization because they are convinced no one else is doing it.
2.5
Current solutions
Google’s Play Store displays a large number of applications under the tag “food waste”. We categorize them according to their main features and provide one example per type: apps providing recipes for leftovers1; apps helping to shop
more efficiently2; apps for sharing shopping lists3, apps that help keep track of food stock and give alerts before the ex-piry dates4; apps that facilitate sharing or donation of food
leftovers5. These apps’ strategies for food waste reduction address mainly the practical reasons, outlined previously, by assisting users with different tasks related to food manage-ment. We found no formal evaluation of their effects but we assume that only persons who already consider food waste a problem would use them.
Another approach is raising awareness of the consequences of food waste. We found a project focused on financial con-sequences, which is still under development by students at the University of Guelph 6. The smart bin they designed is to be placed in a public canteen. It should weigh the food thrown in and display its monetary value, based on the weight and the average price of recently purchased food in the outlet. However, this system has not been evaluated yet. Comber and Thieme [10] designed and tested a system for changing food waste and recycling behaviour through social engagement. BinCam is a bin, equipped with a smart-phone taking a picture every time the lid is open. The item on the image was then recognized (Amazon Turk crowd sourc-ing) and shared on a Facebook application. There, the par-ticipants could monitor group statistics about amount and type of waste, browse the images and participate in different activities. The BinCam increased awareness of behaviour and of wastefulness. Although the participants did not con-sult the Facebook app for others’ behaviours, they did start discussing strategies about food storage and waste disposal among themselves. The authors thus conclude that social engagement is worth further exploration.
2.6
Changing behaviour by influencing
inten-tions
Ajzen [3] argues in his Theory of Planned Behaviour that each behaviour is preceded by a rational intention to
per-1 https://play.google.com/store/apps/details?id= com.lovefoodhatewaste.lovefoodhatewaste 2 https://play.google.com/store/apps/details?id= com.gemoro.afgeprijsd&hl=en 3 https://play.google.com/store/apps/details?id= shopping.list.free.lista.compra.gratis.liston&hl= en 4https://play.google.com/store/apps/details?id=th. co.crie.bestbefore&hl=bg 5https://play.google.com/store/apps/details?id= com.greasedwatermelon.leftoverswap 6http://news.uoguelph.ca/2015/06/ smart-compost-bin-puts-a-dollar-value-on-food-waste/
form it. In short, if we want to change users’ food waste behaviours we should first ensure that they have the inten-tion to minimize their waste. According to the Theory of Planned behaviour, intentions are determined by:
• Attitudes – beliefs about a certain behaviour’s out-comes;
• Subjective norms – beliefs about the social norms (ex-pectations of others), which form a perceived social pressure;
• Perceived Behavioral Control (PBC) – beliefs about factors that impede or facilitate the performance of the behaviour.
Favorable attitude towards the behaviour, perceived social pressure and favorable PBC, would result in a person’s in-tention to perform the target behaviour [3].
2.7
Persuasive technology
Fogg [15] defined the term “captology” - the area where per-suasion and technology overlap. Captology focuses on the design, research and analysis of interactive computer sys-tems created for the purpose of changing people’s attitudes and behaviours. Fogg describes different persuasion strate-gies that can be used depending on the role technology plays. Captology distinguishes among three:
• Tool – technology facilitates the performance of an ac-tion. Such is the approach taken by some of the appli-cations we found in Google’s Play Store – for example, apps like “Best Before” remind the user of a food’s ap-proaching expiry date, making it more likely to have it used on time.
• Media (simulation) – technology provides an experi-ence, allowing users to experience cause – effect rela-tionships. The bin project by University of Guelph, described in chapter 2.4, should help users understand that wasting food is the same as throwing away money. • Social Actor – technology can use the same persuasive techniques humans do. The BinCam system [10] de-scribed in chapter 2.4, allows users to compare their wasting behaviour to the behaviour of other bin users and thus motivates through competition.
3.
PROBLEM STATEMENT AND
RESEARCH QUESTIONS
Although there are a number of practical issues contributing to consumers’ food wasting behaviour, none of them seems to be insurmountable with some effort. At the same time, we assume that persons who do not acknowledge food waste as a problem would not make any effort in the first place. Therefore, we consider it a necessary first step to persuade people of the gravity of the problem and their contribution to it. Our review of related work revealed some gaps this study tries to address:
• Raising awareness about the environmental consequences of food waste - Consumers are often not aware of the
environmental impact of food waste and informing them might stimulate engagement in the issue.
• Raising awareness about users’ actual contribution to the food waste problem - consumers are often not en-gaged with food waste reduction because they do not realize the scale of their own contribution to the prob-lem
• Providing a social norm - food waste behavior is a hid-den process and therefore social norms cannot exercise much pressure. Related work leaves open questions for the potential of providing information about social norms related to food waste reduction.
Our goal is to test the power of raising awareness about en-vironmental impact of household food waste on intentions to reduce food waste as influenced by two factors determining intentions – attitudes and subjective norm [3]. One of the assumptions that we make is that the practical obstacles to consumers’ food waste reduction are rather easy to overcome with simple efforts. Therefore, we will not aim to influence users’ PBC but their perception of the problem, with the assumption that once they realize it, they would feel these efforts are worth it. Following Fogg’s framework about the persuasive roles of computers, we will use simulation to en-hance the effect of awareness information on attitudes and two types of social feedback to affect users’ subjective norms. Therefore, we have three research questions:
RQ 1: Will enhancing awareness about food waste with social feedback, influence people’s intentions to reduce food waste?
Social feedback could influence the effect of awareness about food waste’s environmental impact on users’ intentions to reduce food waste. We expect two possible outcomes, which we will test with the following hypotheses:
H1: Providing users with a descriptive norm about others’ intentions to reduce food waste would increase their own in-tentions to reduce food waste, where it is higher than their original intentions.
H2: Providing users with a descriptive norm about others’ intentions to reduce food waste would decrease their own in-tentions to reduce food waste, where it is lower than their original intentions.
RQ2: Will enhancing awareness about food waste with simulation increase people’s intentions to re-duce food waste?
H3: Simulation can enhance awareness information about the consequences of food waste on the environment by in-creasing its effect on the intentions to reduce food waste. We would also like to compare the strength of each enhance-ment. The third question of this study is:
RQ 3: Is there a significant difference between in-creasing environmental awareness for food waste with social feedback or simulation feedback? If yes, in what ways does their effect differ?
4.
EXPERIMENT
In this section, we describe the prototypes, explain the ex-periment setup and procedure.
We designed two prototypes for applications aiming to influ-ence users’ intentions to reduce food waste. One targeting attitudes and the other subjective norms, as these are the two main factors forming intentions [3]. We then conducted an experiment in order to evaluate the effect of each of the two approaches. The experiment consisted of user testing of the prototypes and questionnaires. With the prototypes, we collected quantitative data about the amount of food waste reduction our participants would be willing to commit to, be-cause of receiving feedback about food waste’s environmen-tal impacts. The questionnaires gathered qualitative data about the factors we aimed to influence – changes in atti-tudes and subjective norms. Although we did not attempt to influence PBC, we did include some questions about ex-ternal factors identified in the literature as influencing food waste behaviour.
4.1
Prototypes
4.1.1
Simulation feedback
One of the reasons consumers do not engage in food waste re-duction is because they think they waste insignificant amounts [24, 29]. According to [25], this is because wasting food hap-pens bit by bit. We aimed at changing this non-engaged attitude with simulation feedback. We applied two of the features of simulation that Fogg [15] describes.
Simulation can provide an experience of a cause-effect re-lationship. In our interfaces, users could see the environ-mental effect of wasting food, at a certain weekly rate ac-cumulated over a longer period. The first simulation was for his/her one-year impact. Afterwards, the user could in-put his/her age and see the impact for his/her remaining “lifetime” (based on the average lifetime expectancy in Bul-garia, 75 years [1]). This intended to let them experience the true scale of the consequences of their food waste, which could not be possible with feedback concerning only one-off wastage.
Our second goal with providing a “lifetime” feedback was to provoke a feeling of moral standards guilt. Moral standards guilt results from violating personal values [4] and it can be used to encourage pro-environmental behaviours [4]. We expected that seeing one’s lifetime “prognosis” could provoke such a feeling of guilt. Due to a desire to compensate for it, we expected users to increase their intentions to avoid wasting food.
Simulation can also help people “rehearse a behaviour” [15, location 855]. Such rehearsal has been found to increase confidence that one can accomplish a certain task or target behaviour. In our case, after seeing their lifetime impact, users could “Try again”. They could then select a differ-ent weekly amount (the time unit selected was one week, as shopping is usually done at least once weekly1) and see
an alternative lifetime impact. We expected that seeing this
1
The weekly waste amounts were based on data for food wasting of bread and bananas in the UK, due to lack of such for Bulgaria.
(a) 1 year (b) ”Lifetime” Figure 1: Simulation for water footprint if wasting 1 banana per week
change of effect would have a significant impact on the en-vironment look feasible.
4.1.2
Social comparison feedback
The social comparison theory holds that people naturally seek to compare themselves to others’ behaviour and atti-tudes, and adjust when forming their own [15]. Our second interface design relied on this theory - the belief that provid-ing users with information about others’ food waste reduc-tion attitudes (social norms) can influence their intenreduc-tions towards food waste reduction.
Social norms for pro-environmental behaviours can provoke feelings of state guilt and can be used to encourage pro-environmental behaviours [4]. State guilt is felt from breach-ing social norms and harms one’s self-esteem, which can lead to reparative actions, i.e. a behaviour change [4]. For exam-ple, if the person finds out she/he were wasting more than the norm, she/he would probably feel state guilt and try to waste less, in order to repair her/his self-esteem. Social norms can also act only as an external motivation [4]. The person complies with the norm only to avoid a feeling of so-cial embarrassment [31]. In this case, the behaviour change would not be sustainable, i.e. as soon as the social pressure disappears, the person might resume his previous behaviour [4]. This is in line with findings described in our literature review – people do not engage in food waste reduction be-cause of lack of social pressure, i.e. they assume others waste as well [17, 33]. Therefore, our design considers two possible effects. If users receive a social norm higher than they cur-rently believe to be normal, their intentions to reduce food waste will increase. A norm lower than currently believed, would decrease their intentions to reduce food waste. Low norms
The interface displayed feedback about environmental im-pact of a piece of wasted food (banana) and enhanced it with
(a) Low norm (b) High norm Figure 2: Social comparison feedback for food waste reduction aims
several low benchmarks. These represented others’ reactions to the same information. The benchmarks were three: the majority of others would not reduce their food waste in or-der to decrease their environmental impact, the rest aim for a low reduction (by 25-50%) and finally, they expect it is “likely” that they achieve their goal. Example is provided in
Figure 2(a). High norms
This app also gave feedback about environmental impact of one piece of wasted food (banana) but enhanced it with high benchmarks. The benchmarks were similar but with the opposite direction: the majority of others would reduce their food waste to improve their environmental impact, the majority of others aim for a high reduction (by 75-100%) and finally, they expect it is “likely” that they achieve their goal. Figure 2(b) shows an example of the high norm interface. In both cases the provided norms were moderate, (i.e. in-stead of “up to 25%”, the low norm was “25-50%”). This was because moderate norms are more likely to provoke as-similation (the individual conforms) and extreme contrast (the individual wants to distinguish her/himself from the group) [11]. The criteria for choosing the social referents, is that they should be similar by a key attribute to the users [11, 15]. We decided on “the other participants in the exper-iment”, as this was the only certain common denominator for all testing users.
4.2
Setup
In order to address the three research questions of the study, we used a mixed between and within-subjects experiment design. Our between-subject factor was presence of enhance-ment and the within-subject was type of enhanceenhance-ment. Our 40 participants were randomly assigned to one of two groups:
• Control Group (20 participants) only received
aware-(a) Feedback for 1 banana (b) Choose a food waste reduc-tion aim
Figure 3: Non-enhanced feedback(control condition)
ness information about the consequences of food waste on the environment.
• Treatment Group (20 participants) received the same information but enhanced once with simulation and once with social feedback.
To counterbalance for learning effects, for the Treatment Group we followed an AB/BA crossover design. The Control Group tested the same combinations and order of awareness information. Appendices A and B provide tables displaying the setup for both groups.
4.2.1
Setting
The experiment was conducted in Sofia, Bulgaria. We chose one of the parks as a setting because the procedure was rather long (20 - 40 minutes depending on the test condition) and people there have some free time.
4.2.2
The “system”
We pretended that the participants work with a fully func-tional system, where the prototypes were connected to a smart bin and an underlying reasoning engine which in fact did not exist. We took this approach as our interests lay only in the interfaces. The other components of the sys-tem were presented as a Wizard of Oz setup, solely for the purpose of evaluation of the interfaces. What we provided was a version of the smart bin, which was equipped with a camera and a device that supposedly communicated with the interface. We also provided participants with a tablet (Lenovo A10-70 A7600) on which our prototype interface was available in a functional version. All participants re-ceived information about the environmental impact of food waste on the tablet, after throwing some food in the ”smart” bin. They were lead to believe that the tablet is connected to the “smart” bin, which can recognize the type of food by making use of the camera, which sends the image to the
Figure 4: Setting.
application on the tablet, so that an appropriate feedback could be provided. The main role of the bin was to help participants imagine that it is possible to receive feedback related precisely to their personal waste.
4.2.3
The opportune moment
Asking users to throw food into a bin aimed at creating an opportune feedback moment. Sending feedback right after throwing food should have the increased effect of the “kairo” principle, as described by [15], as directly after throwing away food people (are expected to) feel uncomfortable be-cause they did something wrong.
4.2.4
The food thrown and awareness content
Each participant had to throw away four items of food - two bananas and two pieces of bread. We selected these food types based on two criteria. First, it had to be food that most people waste. Due to lack of such data about Bulgaria, we based our choice on a study in the UK, according to which bread and banana are the most wasted food types [28], and on statistics that bread and fruits are among the most consumed foods in Bulgaria [2].
The second criteria concerned the delivered information. All participants received information about two aspects of the environmental impact of food waste – its carbon [7] and water footprints [22]. As the size of both footprints varies among products, we wanted to choose one product with a low and one with a high impact. In our case, banana has lower carbon and water footprints then bread. This allowed us to see to what extent value has an effect on intentions.
4.2.5
Procedure
Before taking part, the participants completed a short ques-tionnaire collecting demographic information and assessing their current behaviour and attitudes towards food waste. The main goal of this questionnaire was to detect possible history threats and to filter our target population - only per-sons who did admit to wasting food and were aged 18 - 65 were included in the study. We took into consideration that people above 65 almost do not waste food [24].
We had users test the two prototypes shown in Figure 1 and Figure 2. A control group received the same awareness information with no enhancement(Figure 3). To test the prototypes both groups had to perform four sets of tasks (dispose four pieces of food), following the same procedure: I. State their current beliefs about the water/ carbon foot-print of banana/ bread on the environment.
II. Throw a piece of bread/ a banana in a “smart” bin. After this, they received awareness information about the specific food item on the tablet. Depending on the condition, it was either supported by social comparison or by simulation feed-back or non - enhanced.
III.State whether, as a result of receiving the information, they would like to reduce their food waste for the thrown type of food or not.
IV. In case they stated a positive intention, they were prompted to choose by what percentage.
V. State how confident they feel about following their aim. (Appendix E shows an information flow diagram for the three conditions.)
Once the participants did a cycle, they had to fill in a ques-tionnaire related to the tested feedback type. The question-naire assessed the three determinants of intention [3] - their attitudes towards food waste reduction, subjective norms and perceived behavioural control. Once finished with the experiment, they were thanked and received a small pack of plums1.
5.
RESULTS AND ANALYSIS
5.1
Population
We recruited 45 participants, out of which 40 were included in the final sample. Our population was mostly young – 36 were aged 18 – 45 and 4 were above 45. Both groups were predominantly female – 11 in the Treatment group and 16 in the Control group. Twenty had children (8 in the Control group and 12 in the Treatment group). They were relatively highly educated (24 had a master’s degree or higher, 7 bach-elor degree and 8 high school) and all but 3 had an income above the average for Bulgaria (only 2 received the minimum wage; 1 had no income).
Most of the participants we can consider low food wasters for Bulgaria [9] - 21 admitted to waste 0,8 kg weekly or less, 15 fall in the range of 0,8 – 1,6 kg and only 3 waste more than 1,6 kg per week. All 40 participants agreed that they could reduce their food waste if they made an effort. Concerning their attitude towards global warming – 2 did not answer, 25 disagree that the human factor is exagger-ated, 8 remained neural and 5 agreed (five participants who strongly agreed, were not included in the final sample, with the assumption that they could not be possibly influenced by the information about carbon footprint).
1
This refers to a popular Bulgarian fairy tale called ”Plums for waste” which we also used as a title to announce the experiment. In the story, the king offers plums in exchange for household waste. The person who brought in the least garbage received a greater reward - to marry his daughter.
5.1.1
Lack of Awareness
Overall, our population was environmentally conscious – all participants stated that they try to consider the environ-mental consequences of their actions. However, more than half (23) agreed or strongly agreed with “Food waste has no negative impact on the environment because it is nat-ural and biodegradable” and only 13 disagreed or strongly disagreed, 4 remained neutral.
Additionally, all participants were prompted to estimate the sizes of water and carbon footprints for banana and bread. Concerning the water footprint, estimations were close (ta-ble provided in the Appendix D), but informal conversations showed that no one was aware of the notion of “grey wa-ter” [22] and most considered water simply to be used for irrigation. They were asked to compare the food’s footprint to a car’s footprint - “Try to estimate the distance that a car needs to travel so that it cases the same damage as a ba-nana/ piece of bread”. Their guesses about carbon footprints show that they do not have any intuition of the problem (ta-bles are provided in the Appendix C). These results are in line with previous studies stating that people are unaware and often mistaken about the connection between climate change and food waste [24, 29].
5.2
Quantitative data
We compared the aims for food waste reduction of the par-ticipants in the two groups – control group receiving only awareness information and the treatment group, receiving the same information but enhanced with social information and simulation. We measured the size of intentions on the following scale: reduction by 0-25%; 25 – 50%, 50-75%, 75-100%, 100%. For the analysis, each interval answer was as-signed to the upper value of the range. Persons who stated they do not want to reduce their waste, were assumed to have 0% reduction aims. This assumption should be taken into account when considering the provided results. Positive effect on intentions
All three conditions – control, social feedback, simulation feedback - provoked a desire to reduce food waste. All 40 participants stated that based on the received information, they would like to reduce their food waste for at least one of the tasks. Moreover, for all cases the average stated aims were rather high - above 70% reduction (Figure 5). The “No” food waste reduction cases
There were several cases, in which the participants stated they would not like to reduce their food waste. Although these were few, we found it important to look for a possible explanation to the lack of effect of the provided information. Two persons in the Control group stated that they would not like to reduce their food waste on one or more tasks. One participant’s “No” answers were related to the two tasks re-lated to bananas. It is important to note also that this same participant stated she/he almost never wastes bananas. The other person stated “No” for three tasks – carbon footprint of bread, water and carbon footprints of banana. In the Treatment group, 3 stated “No” for one of the social feed-back tasks (throwing banana and receiving feedfeed-back for its carbon footprint) and 1 for a simulation task (wasting a ba-nana and receiving feedback for its water footprint).
Figure 5: Food waste reduction aims for the different con-ditions.
From the above, it appears that feedback for banana was less influential. This is in line with our expectations, as both banana footprints are lower than footprints of bread. However, comparing all food waste reduction aims for tasks related to each food type revealed the difference is small – mean 85% +/- 27.2, as opposed to 83% +/- 32 for bananas. Wilcoxon Signed Rank Test confirmed it is insignificant (p = 0.5).
5.2.1
Testing hypotheses
The results from our quantitative data show that we have no sufficient evidence that either social comparison feedback or simulation feedback can affect consumers’ intentions to reduce food waste. For brevity, we report only the outcomes and leave here for reference Figure 5 showing the changes in intentions in the different conditions. The complete results can be found in Appendix F.
H1: Effect of Comparison with high social norms
Nevertheless, the aims set for high social norm feedback were higher than the aims stated for the same information with-out enhancement, as can be seen in Figure5, H1. A Mann Whitney U test was run and determined that this difference is statistically insignificant (p = 0.7).
H2: Effect of comparison with low social norms
On the other hand, the data indicates that providing social norms could also have a negative effect, i.e. reduction aims were lower for low social norm feedback responses, as op-posed to the corresponding control group aims (Figure5, H2). It should also be noted that under this condition the greatest number of persons stated they would not like to re-duce their food waste (3/20). However, a Mann Whitney U test determined that the decrease is statistically insignificant (p = 0.2).
H3: Effect of Simulation
Similarly, the results do show an increase in aims for food waste reduction when comparing simulation feedback to the control condition (Figure5, H3). However, the Mann Whit-ney U Test we ran showed that the increase is statistically insignificant again (p=0.6).
Figure 6: Attitude change across the three conditions.
Social comparison vs simulation feedback
When comparing the two enhancements, high social norms feedback appears to have had stronger effect on reduction aims than simulation (Figure5). In this case again, the increase is statistically insignificant - Wilcoxon Signed Rank Test, p = 0.5.
Although none of our hypothesis was confirmed, we have indications supporting each of them. We will next review the questionnaire responses for possible strengths and weak-nesses of each prototype.
5.3
Qualitative data
We collected qualitative information about factors identi-fied in the literature as influencing intentions to reduce food waste and about the way the prototypes were perceived. Be-low, we examine each factor separately and look for possible connections to the quantitative data. Unless stated other-wise, for all of the questions discussed, the scale was Likert, 5 points, coded as 1= Strongly Agree, 2=Agree, 3 = Neutral, 4 = Disagree and 5 = Strongly disagree.
5.3.1
Attitude changes
After testing the prototypes, all participants answered three questions about attitudes change in the three conditions. The scale was positive, i.e. agreement showed a positive effect.
The set inquired: to what extent the presented information clarified the connection between food waste and its water footprint, food waste and its carbon footprint, and to what extent the gravity of the environmental impact of food waste was made clear. As can be seen in Figure6, the partici-pants largely agreed with all three statements. In order to
Figure 7: Perceived effect of simulation, stronger than social comparison.
compare the effect of each type of feedback, we took the mean of all responses to each question. The mean of these means we assume to represent the attitude change effect. The means show slightly better results for the control condi-tion than for both simulacondi-tion and social feedback(Figure6). The differences are very small thus we could only state that according to this measure, there is positive attitude change regardless of the condition.
5.3.2
Simulation perceived as having a stronger
ef-fect than social comparison
Participants from the Treatment group were also asked to what extent the information specific to each of the enhance-ments (information about others’ reduction aims and infor-mation about their own food waste’ environmental impact over time) helped them decide about the waste reduction aims they set. The responses for this question show an ad-vantage for simulation over social comparison as can be seen on Figure7 – 19/20 participants agreed that simulation in-formation was helpful and 1/20 remained neutral; as op-posed to 9/20 agreeing and 5/20 disagreeing; 6/20 neutral. The difference is significant as a Sign test showed (p = 0.02).
5.3.3
Simulation
Food waste environmental impacts over time
The main feature of the simulation feedback was present-ing our participants their food waste environmental impact as accumulating over time. The assumption was this would help them to understand the scale of their contribution to the problem. The participants were led through three steps: first receiving information about the food they throw away, afterwards they could opt to see information about its im-pact (if wasted) for one year, and, last, information about impact for their remaining ”lifetime”. We wanted to know which one had the strongest effect. According to 12/20 participants information for one year had a stronger effect on them than information for one item and the “lifetime” simulation was more effective than the 1 year simulation (Figure8). This positive response might be the explana-tion to the perceived stronger effect of simulaexplana-tion over social feedback on their intentions.
Experimenting with different amounts
Users also had the option to try the change of impact by selecting different weekly amounts of waste. We were ex-pecting this would help them make a better informed de-cision and feel more confident about following it through. However, only 3/20 participants actually tried this option.
Figure 8: State and moral standards guilt.
5.3.4
Social comparison
Perceived effects of low social norm
One of our hypotheses was that providing a low social norm might decrease intentions for food waste reduction by lower-ing the perceived social pressure. However, to the statement “I would feel I have accomplished my duty in case I follow a food waste reduction goal at least as high as the goal of oth-ers” only three participants agreed, 16/20 disagreed and one remained neutral. This strong negation is contradictory to our quantitative results, which showed lower aims (although insignificantly) connected to low social norm(Figure8). Moral standards and state guilt As food wasting is usually associated with feelings of guilt [24, 29], we wanted to see whether simulation or social treatments could enhance this feeling with the aim of stimulating food waste reduction intentions. In the pre-test, almost all 34/40 participants agreed that they feel guilty when wasting food and 5/40 admitted they do not; 1/40 remained neutral.
However, neither simulation, nor social feedback elicited a strong feeling of guilt. For simulation, only 3/20 participants in the Treatment group, agreed and 14/20 disagreed that if their lifetime impact was large, they felt guilty(Figure8). For social comparison, 7/20 agreed and 9/20 disagreed that they felt guilty to aim for a lower food waste reduction amount than the others(Figure8). Previous studies sug-gesting that guilt is associated with food waste were con-firmed [24, 29]. Nevertheless, influencing moral standards guilt (felt from breaking internal values) was not successful with the simulation feedback we provided. Social feedback, which we expected to affect state guilt (felt when breaking social norms), elicited stronger results.
5.3.5
Perceived Behavioral Control (PBC)
Following [3], we asked all our participants about their PBC, i.e. to what extent they think reducing food waste is a fea-sible task.
Confidence
For each task, participants had to state how likely it is that they achieve their food waste reduction aims. In all cases, they showed a high confidence, with a small advantage for social comparison condition (Figure9).However, the differ-ences are insignificant(Wilcoxon Signed Rank Test, p=0.966; Mann Whitney U, p = 0.705). Considering the experiment design, we interpret these results only as an encouraging in-dication that none of the participants saw serious obstacles to achieve their aims. In order to assess the validity of the stated confidence, users’ future behaviour would need to be
Figure 9: Confidence responses for all 4 tasks across the three conditions.
observed. External factors
Participants were also asked to what extent they feel they are responsible for their food waste and its consequences on the environment. The Lyndhurst [24] survey showed that one reason people do not engage in food waste reduction is that it is not a priority. So, our first question was “I will not be able to engage in food waste reduction, as I have other more important worries”. Lindenberg and Steg [23] point out that people can shift their own responsibility towards a pro-environmental behaviour to larger actors. Therefore, our second question was “The contribution of industries to global warming would make my own efforts for food waste reduction meaningless”. To both statements, all participants disagreed and only one admitted that food waste is not a priority of hers. Although this is good news, we did not ask questions related to practical obstacles for food waste reduction as our prototypes were not addressing these. It is possible that such factors play a significant role in reality.
5.3.6
Preferences
All participants but one in the control condition agreed that they would like to use such an application if available. In the treatment group, there was one person who would not like to use simulation. It should be noted, that this same participant did not actually see the simulation information but opted to skip it.
How long will the effect last
According to our participants, the provided awareness infor-mation will have a lasting effect. About half of the responses for all three conditions were that it will influence their at-titude towards bananas and bread “For my whole life” (8 of 17 for simulation (Treatment group); 9 out of 18 for so-cial comparison(Treatment group) and 9 out of 19 for the control condition). For social feedback, the range for the re-maining half was from “1 week” to “More than 1 month”, for simulation was “1 month” and “More than 1 month”. When participants in the treatment group were asked to rate the simulation and social feedback on the expected longevity of their effect (on a scale from 1 to 5, with 5 = longest effect), simulation also shows a small and statistically insignificant (Wilcoxon Signed Rank Test, p=0.151) advantage. For sim-ulation the mean = 4.1053 (std. dev = 1.04853 min = 2, max = 5); for social feedback mean = 3.6316 (std. dev = 1.6479, min = 2, max = 5). This is however, merely a one-time indication that could only serve as a possible direction for future work.
More or less
For all three conditions, participants mostly agreed that the more often they receive such feedback, the more likely it is that they will attempt to reduce their food waste (20 agreed in the control group; 16 agreed or strongly agreed and 1 disagreed in the simulation condition; and 16 agreed or strongly agreed and 3 disagreed in the social feedback con-dition). Nevertheless, a following question revealed that the preferred frequency of such messages varied greatly among individuals and not among conditions. The most frequent answer in the control group was “Once per week” (6 per-sons). Whereas, for both simulation and social feedback, 5 stated they would like to receive such messages “Every time they waste food”; 5 opted for “Once per type of food” and 5 stated they would like to receive it “Once monthly”. As can be seen, the answers are quite heterogeneous. This might indicate that such information is not always welcome and should not be forced upon users, but rather be adjusted to their personal preferences.
6.
DISCUSSION
We derive the main findings of this research and take its limitations into account. The most important is that we only measured stated intentions for food waste reduction, with the assumption that more serious intentions lead to larger behaviour change [34]. In order to claim validity, we would need to trace whether the participants actually started wasting less food.
6.1
Raising awareness
Our results show that information about the environmental impact of food waste did influence participants’ intentions to reduce food waste but this was independent of the con-ditions. For all three conditions aiming to motivate users, the stated reduction aims were high – 80%. Moreover, all but one participant felt confident of being able to achieve them. Considering that our population was largely unaware of both water and carbon footprints of food waste, it seems natural that providing pure information caused a strong re-action. Moreover, they agreed that the information made the connection between food waste and its environmental impact clear. Our findings support [13]’s that environmen-tal concerns can be a motivation for food waste reduction. It is worth following to what extent such information leads to actual change of behaviour after a possible novelty effect [8] has passed. We should also note that a social desirability effect [16] cannot be excluded, especially since all partici-pants admitted to generally feeling guilty for throwing away food.
6.2
Simulation
The simulation condition induced higher food waste reduc-tion aims than the control condireduc-tion. Although the differ-ence is insignificant, the questionnaire responses point out that one feature of the simulation might have contributed to this effect. Participants stated that seeing long-term im-pacts helped them decide on their reduction aims and influ-enced them more than the impact of only one wasted item. Lifetime was perceived as more influential than one year. Therefore, we assume that helping users realize the scale of their environmental impact does have a potential to moti-vate food waste reduction behaviour.
Besides seeing their future environmental impact, users could also see what they can do about it - to what extent they could affect it, if they change the amount they waste. This feature was disappointingly neglected - only three people ac-tually tried it. This suggests that providing users with tai-lored information is a better strategy than expecting them to make an effort to look for it. It is also possible that par-ticipants did not make a conscious choice but simply did not notice the “Try again” button. This we infer because some of them asked about the meaning of question referring to this function.
6.3
Social comparison
High social norm provoked the highest reduction goals com-pared with the two other conditions. On the other hand, when provided with a low norm, participants set the lowest goals. Although the differences were insignificant in both cases, we believe that they give an important indication -the social norms we provided only acted as an external mo-tivation. When the social pressure was lowered, the inten-tions also reduced. Therefore, even the encouraging results of high norm feedback should be considered with care, as their effect might be only temporary [30].
Feelings of social guilt provoked by others potentially hav-ing higher food waste was the main feature we expected to motivate food waste reduction. Fewer than half of the par-ticipants admitted to feeling such guilt. The questionnaire responses did not confirm the low norm results either. Al-most all participants disagreed that they would feel they had done “their bit” as long as they do not aim lower than the rest. Although these responses are contradictory to the quantitative results, they are not unexplainable. It is possi-ble that participants were more influenced by social norms than they would admit, since “participants deny conformity as an explanation for their behaviour” [14, p. 246].
6.4
Simulation or social comparison?
When comparing the two enhancements, there appears to be a contradiction between our qualitative and quantitative results. If we look at the reduction aims, participants seem more influenced by the social comparison feedback but they self-reported to be affected stronger by simulation. This difference in perception could be attributed to two charac-teristics of the study. First, the newly received factual infor-mation was apparently impressive. As simulation is directly resultant from it, it is possible that a novelty effect enhanced the way it was perceived. Measuring the perceptions after some time of using the prototypes, might reveal a diminished effect of the simulation. Second, as mentioned in section 6.3, it is possible that participants were more influenced by so-cial norms than they self-reported. It is worth noting that simulation could be the more sustainable approach towards behaviour change - our participants perceived it as poten-tially having a longer lasting effect than social comparison. These answers are not a surprise - they are in line with other research suggesting that internal motivation could lead to a more stable change of behaviour than external motivation [4]. In our case, simulation aimed to affect our participants’ moral standards, so it can be considered internal motiva-tion [4]. The provided social norms we categorize as an ex-ternal motivation. Although it is possible that they become
internalized [18], we believe this was not the case with our experiment.
6.5
Guilt and food waste reduction
Our attempt to invoke moral standards and social guilt seems to be unsuccessful, despite the fact that all partici-pants did agree to feeling guilty about wasting food. One possible explanation could be that they did not actually feel responsible for wasting the food while testing the prototypes as they followed a scenario. Guilt usually increases when the person feels in control of the impact of her/his actions be-cause this implies responsibility for these actions [6].
7.
CONCLUSION AND FUTURE WORK
We designed and tested two prototypes of interfaces aim-ing to influence users’ intentions to reduce their food waste. Both presented users with information about food waste’s impact on the environment, one with simulation and the other with social comparison feedback. The purpose of our experiment was to assess whether either of these two ap-proaches holds a potential for behaviour change.
We have no sufficient quantitative evidence that enhancing environmental awareness information with either social or simulation feedback can have an effect on intentions to re-duce food waste. Participants did aim for a greater reduc-tion percentage with simulareduc-tion and positive social feedback when compared to the control condition. Despite this, their effect seems to have been less salient than the effect of raising awareness, as in all cases the aims for food waste reduction users set were relatively high in all three conditions. Sup-porting previous research, participants were unaware about the environmental impact of their food waste. Once they realized it, they declared a strong desire to avoid it. Among the two prototypes, users preferred simulation over social comparison. Information about food waste’s impact accumulated over time was perceived as significantly more useful when deciding on one’s aims for food waste reduction, than information about others’ food waste reduction inten-tions. Furthermore, simulation for a “lifetime” was reported to be more influential than for one year. The participants did not admit to conforming to social norms, however we have indications that they do affect them - social pressure stimulated the higher food waste reduction goals in our ex-periment but also the lowest when the norm was low. This in line with literature suggesting external motivations do not provide a stable behaviour change [30].
In conclusion, although we have to reject our hypotheses that simulation and social comparison can influence food waste reduction, we find both worthy of future work. Simu-lation mainly because it was well accepted. Social compar-ison should be considered with care in view of its possible negative effects.
7.1
Future work
For future research we recommend the following:
• Inclusion of a phase in which the following behaviour is measured, not only stated intentions.
• The provided feedback might elicit different results if it were related to actual food wasting by the participants, instead of them following a scenario.
• Additionally, if the prototypes were tested over time, the experienced effects might be different as the nov-elty effect of the awareness information would dimin-ish.
• Giving feedback on user’s behaviour change in reality and its lifetime impact might be more motivating than our one-off study.
• Future studies should consider using known peers as social referents. According to [11], the effect of social comparison feedback might be greater in both negative and positive directions.
8.
ACKNOWLEDGEMENTS
I would like to thank my supervisor Frank Nack for the sup-port, detailed feedback and understanding to my situation. I am also very grateful to Lynda Hardman for agreeing to become second reader on such a short notice. Special thanks go to Svetlin Stoev as well, for helping with the prototypes and to Florian Amersdorffer for crafting the “smart” bin. Thank you to all participants.
References
[1] National statistical institute - lifetime expectancy for bulgaria, 2013.
[2] Report of the bulgarian national statistics insti-tute for household incomes, expenses and consump-tion. http://www.nsi.bg/sites/default/files/ files/pressreleases/HBS2014_GO8KJYP.pdf, 2014. [3] I. Ajzen. The theory of planned behavior.
Orga-nizational Behavior and Human Decision Processes, 50(2):179 – 211, 1991.
[4] T. Bedford, P. Collingwood, A. Darnton, D. Evans, B. Getersleben, W. Abrahamse, and T. Jackson. Guilt: an effective motivator for pro-environmental behaviour change? Working paper, University of Surrey, 2011. [5] A. Bektes. Research and product design to minimize
food waste in Western domestic kitchens. PhD thesis, TU Delft, Delft University of Technology, 2010. [6] M. Berndsen and A. S. Manstead. On the relationship
between responsibility and guilt: antecedent appraisal or elaborated appraisal? European Journal of Social Psychology, 37(4):774–792, 2007.
[7] M. Berners-Lee. How bad are bananas. The carbon footprint of everything. London: Green Profile, 2010. [8] G. H. Bracht and G. V. Glass. The external validity
of experiments. American educational research journal, pages 437–474, 1968.
[9] K.-R. Br¨autigam, J. J¨orissen, and C. Priefer. The ex-tent of food waste generation across eu-27: Different calculation methods and the reliability of their results. Waste Management & Research, 32(8):683–694, 2014.
[10] R. Comber, A. Thieme, A. Rafiev, N. Taylor, N. Kr¨amer, and P. Olivier. Bincam: designing for engagement with facebook for behavior change. In Human-Computer Interaction–INTERACT 2013, pages 99–115. Springer, 2013.
[11] K. Corcoran, J. Crusius, and T. Mussweiler. Social com-parison: Motives, standards, and mechanisms. Theories in social psychology, pages 119–139, 2011.
[12] J. Cox, S. Giorgi, V. Sharp, K. Strange, D. C. Wil-son, and N. Blakey. Household waste prevention—a review of evidence. Waste Management & Research, 28(3):193–219, 2010.
[13] N. Doron. Waste not, want not: How fairness concerns can shift attitudes to food waste, 2012.
[14] M. Eysenck. Psychology for AS Level. Psychology Press, 2005.
[15] B. Fogg. Persuasive technology: Using computers to change what we think and do (interactive technologies). 2002.
[16] A. Furnham. Response bias, social desirability and dissimulation. Personality and individual differences, 7(3):385–400, 1986.
[17] E. Graham-Rowe, D. C. Jessop, and P. Sparks. Iden-tifying motivations and barriers to minimising house-hold food waste. Resources, Conservation and Recy-cling, 84:15–23, 2014.
[18] V. Griskevicius, R. B. Cialdini, and N. J. Goldstein. Social norms: An underestimated and underemployed lever for managing climate change. International Jour-nal of Sustainability Communication, 3(1):5–13, 2008. [19] D. Gunders. Wasted: How america is losing up to 40
percent of its food from farm to fork to landfill. Natural Resources Defense Council Issue Paper. August. This report was made possible through the generous support of The California Endowment, 2012.
[20] J. Gustavsson, C. Cederberg, U. Sonesson, R. Van Ot-terdijk, and A. Meybeck. Global food losses and food waste. Food and Agriculture Organization of the United Nations, Rom, 2011.
[21] M. Halstead, T. Kober, and B. van der Zwaan. Under-standing the energy water nexus, 2014.
[22] A. Y. Hoekstra. The water footprint of food. 2008. [23] S. Lindenberg and L. Steg. Normative, gain and hedonic
goal frames guiding environmental behavior. Journal of Social Issues, 63(1):117–137, 2007.
[24] B. Lyndhurst. Food behaviour consumer re-search—findings from the quantitative survey. Briefing Paper. UK: WRAP, 2007.
[25] L. McKnight-Yeates. Out of sight, out of mind: what influences our perception of waste and activates our in-tention to live more sustainably? PhD thesis, Royal Roads University, 2009.
[26] J. Parfitt, M. Barthel, and S. Macnaughton. Food waste within food supply chains: quantification and potential for change to 2050. Philosophical Transactions of the Royal Society B: Biological Sciences, 365(1554):3065– 3081, 2010.
[27] T. Quested, E. Marsh, D. Stunell, and A. Parry. Spaghetti soup: The complex world of food waste behaviours. Resources, Conservation and Recycling, 79:43–51, 2013.
[28] T. Quested and L. Murphy. Household food and drink waste: A product focus, a report for warp, 2014. [29] E. M. Research. Wrap. we don’t waste food! a
house-holder survey, 2007.
[30] R. M. Ryan and E. L. Deci. Intrinsic and extrinsic mo-tivations: Classic definitions and new directions. Con-temporary educational psychology, 25(1):54–67, 2000. [31] T. J. Scheff. Shame and the social bond: A sociological
theory. Sociological Theory, 18(1):84–99, 2000. [32] R. Th¨onissen. Food waste: The netherlands. In
Presen-tation to the EU Presidency Climate Smart Food Conf, 2009.
[33] P. Tucker and P. Douglas. Understanding household waste prevention behaviour. Final report. WR0112, 2007.
[34] T. L. Webb and P. Sheeran. Does changing behavioral intentions engender behavior change? a meta-analysis of the experimental evidence. Psychological bulletin, 132(2):249, 2006.
Appendix A. Experiment setup for the Treatment group
A = Social comparison, B = Simulation;
H = high, L= low;
ln = low social norm, hn = high social norm
Participant Level Content Food 1 Level Content Food 2 Level Content Food 3 Level Content Food 4
1-5 A H
/ hn
C02 bread L
/ ln
H20 banana B H H20 bread L C02 banana
6 -10 A L
/ ln
C02 banana H
/hn
H20 bread B L H20 banana H C02 bread
11-15 B H H20 bread L C02 banana A H /hn C02 bread L / ln H20 banana 16 -20 B L H20 banana H C02 bread A L / ln C02 banana H /hn H20 bread
Appendix B. Experiment setup for the Control group
C = Control condition
H = high, L= low;
Participant Level Content Food1 Level Content Food 2
Level Content Food 3 Level Content Food
1 - 5 C H C02 bread L H20 bana
na
H H20 bread L C02 banana
6 - 10 C L C02 banana H H20 bread L H20 banana H C02 bread
11 - 11 C H H20 bread L C02 bana
na
H C02 bread L H20 banana
16 - 20 C L H20 banana H C02 bread L C02 banana H H20 bread
Appendix C. Estimations for carbon footprint if compared to a car’s footprint driven for a certain
distance (meters)
Bananas Bread
mean 72146 283170
SD 230182 1580121
true value 114 170
Appendix D. Estimations for water footprint (liters)
Bananas Bread
mean 56 58
SD 52 173
