Changing Consumer Behavior to Decrease Food Waste

Changing Consumer Behavior to Decrease Food Waste

Thesis Design for Information Studies (HCM)

Bart Witteveen (10581243)

Bart Witteveen

bart.witteveen@students.uva.nl

Bart Witteveen

bart.witteveen@students.uva.nl

Supervisor: Frank Nack (University of Amsterdam) 2nd Reviewer: Robbert-Jan Beun (Utrecht University)

Changing Consumer Behavior to Decrease Food Waste.

ABSTRACT

Food waste is a serious problem that consumes a very large proportion of various resources such as energy, land and drinking water. In the developed world, a lot of food is wasted in people’s homes for simple reasons. This means it is easily avoidable by changing attitudes and behavior. In this report an application is presented that aims at changing the behavior of wasting food by allowing users to reflect on their behavior and motivat-ing them to change it. The application was tested by five households during a two weeks experiment. All of the users of the system said they were influenced by the system and some of them made serious changes in their behavior. This report describes the design of the application and the results of the experiment.

General Terms

Behavior, Captology, Mobile, Persuasive Technology, Self-Monitoring.

Keywords

Application Design, Awareness, Attitude Change, Be-havior Change, Diary, Food Waste, Qualitative Study.

1.

INTRODUCTION

In the past 50 years the total production of food has seen a large increase, varying between different food products and animals from around 40% up to 350% (Godfray et al., 2010). However, this same research shows that roughly 30 to 40 percent of all food is wasted. It is not only the food itself that is wasted, but also the resources that were used to produce that food. Some facts by Gunders (2012) give an impression of the scale of the problem in the US:

• Getting food to consumers takes around 10 percent of US total energy consumption.

• 50 percent of US land is used for food production. • 80 percent of US freshwater consumption is used

for food production.

• Reducing waste by 15 percent could feed 25 million people every year in the US.

The seriousness of the problem is backed up by research. WRAP (2008) estimates that food waste costs the UK over 10 billion per year and Van Westerhoven and Steen-huizen (2010) estimate that 72,9 kilograms of food is wasted per person every year in the Netherlands, of which 43,7 kilograms was avoidable.

Where in the food chain this waste takes place varies for different areas in the world. In the developing world most of the waste takes place in harvesting, processing and transporting food due to a lack of facilities, such as cold storage or pest control (Godfray et al., 2010). In developed countries, however, a very large part of the food is wasted at people’s homes (Godfray et al., 2010). Since most of the food waste in the developed world is not caused by a lack of facilities (e.g. fridges), we can conclude that it is caused by the behavior of people. Therefore, the goal of this research is to design a mobile application that changes this behavior. Five households were asked to test it for two weeks to find out whether it makes them think of the problem of their personal food waste and address it by changing their behavior. The next section will give a more detailed overview of the food waste problem and its causes based on available literature and research. In addition, some theory on behavior change will be discussed. After this literature review, I will define the problem statement. Section four, five and six will be about the design of the system, the setup of the test and the test results respectively. In the last two sections I will conclude and suggest future work.

2.

RELATED WORK

2.1

Waste Numbers and Percentages

Different research reports about food waste give a good overview of why, when and how food is wasted. Van West-erhoven and Steenhuizen (2010) have studied food waste in the Netherlands and WRAP (2008) has done the same in the UK. Both have the same approach; they combine the inspection of waste with the interview-ing of people. The main difference is the size of the study; Van Westerhoven and Steenhuizen (2010) have inspected waste of 110 households, whereas WRAP (2008) had a significantly larger study with 2138 households. Van Westerhoven and Steenhuizen (2010) estimate the waste at 72,9 kg per person per year in the Nether-lands. The amount WRAP (2008) estimates for the UK is nearly the same at 70 kg. Both studies take into account unavoidable food waste, which consists of food that is not eatable, such as peels, stumps, teabags, etc. Except for this type of waste, most of the waste in both studies is (possibly) avoidable. The percent-age of avoidable food waste is similar for both studies; Van Westerhoven and Steenhuizen (2010) estimate that for the Netherlands this percentage is 58% and WRAP (2008) estimates that for the UK it is 61%. The total costs of all this waste is estimated at £420 per year for an average household in the UK (WRAP, 2008).

It turns out that in the UK nearly a quarter of food is thrown away as a whole or unopened (WRAP, 2008). In the Netherlands 10 percent of the food is wasted unopened (as a whole is not counted) (Van Westerhoven and Steenhuizen, 2010). Another significant percentage of the waste is that of prepared food, in the UK and the Netherlands this percentage is estimated at 27 and 28 percent respectively. Both of these categories of waste are relatively easy to prevent.

As regards to the composition of the avoidable waste, the UK and Dutch studies also show similarities. Both studies show that vegetables, bakery, fruits are thrown away in large quantities, combining for around 50% of the waste (assuming that WRAP (2008) counts pota-toes as vegetables) (WRAP, 2008; Van Westerhoven and Steenhuizen, 2010). Other products that combine for a large percentage of the waste are: mixed foods, drinks and fish & meat (WRAP, 2008; Van Wester-hoven and Steenhuizen, 2010). The main difference is that in the Netherlands a lot more diary products are wasted (12% of the waste compared to 4,6% in the UK) (WRAP, 2008; Van Westerhoven and Steenhuizen, 2010). This can be explained by the fact that diary products form a substantial part of the Dutch eating culture.

2.2

Behavior and Causes

To change the behavior of people so that they will waste less food, it makes sense to have a look at why they are wasting food in the first place. This section describes different causes of food waste and gives and overview of the current behavior of people when it comes to food waste.

In a study by Lyndhurst (2007) about food behavior 1862 UK households were interviewed. One of the sub-jects of this study was the reason that people throw away food. The most important reasons were (in de-scending order):

• Food gone past its use by or best before date. • Tempted by special offers.

• Food visibly gone bad or smells bad. • Made too much food.

• Not eating the foods that need to be eaten first.

None of these are true problems that required some kind of innovative solution in terms of technology. All of them, theoretically, could be avoided only by paying a bit more attention to food waste and trying to prevent it. And that is exactly where the problem sits; attention for food waste seems to be underdeveloped. Some re-sults of the study by Van Westerhoven and Steenhuizen

(2010) clearly confirm this, since they show that people who put a little more effort in preventing food waste clearly waste a lot less. The results were:

• People who agreed with the statement that they try their best to reduce food waste waste 14 kg less than people who were neutral to this statement and 26 kg less than people who disagree with the statement.

• People who throw away food immediately when it is passed its “best before” date waste nearly twice as much as people who do not do this (56 kg vs 31 kg)

• People who say they try not to buy too much food waste waste 6 kg less than people who sometimes do this and 27 kg less than people who never do this.

Simply put, it is the small efforts that can make a really big difference in the amount of food wasted. Everyone has the ability to decide to eat something first when it is about to go passed its expiration date or buy less food in the shop, but apparently most people are not bothered to do it. The survey by Lyndhurst (2007) explains why this is the case. Their results show that the two main reasons mentioned by people who are not bothered by food waste are that they think they do not throw away much food and that they do not consider it a problem. Gunders (2012) adds that one of the drivers for food waste is the undervaluing of food and a lack of awareness of the food waste problem. The system will focus on these points and tries to motivate people to start doing these small things to reduce the amount of food wasted.

2.3

Existing Applications

A lot of applications for shopping and food management already exist. A simple search in the Google Play store on “shopping list” gives over 250 results, some of them being installed millions of times. Applications giving insight in food dates and products are also around, but in smaller numbers.

Two examples of both categories include “Best Before”

1_{and “Out of Milk Shopping List”}2_{. The former allows}

users to add products to their inventory by scanning the bar code or adding them manually. A list shows the products sorted by their expiration date and users can set notifications to be reminded when a product is close to its expiration date. “Out of Milk” allows the user to make shopping lists. they can also keep track of their inventory, but the application does not support expiration dates.

1_{http://goo.gl/CxNQGf} 2

Another system that can help prevent food waste is “Pervasive Frigde” by Rouillard (2012). It is based on the same idea as “Best Before”, but explained in a full research report. It focuses on the connection with a product database and notifications that the user gets. In addition, users can use different ways of data in-put, such as image/voice recognition, bar code scanner or traditional keyboard input. It uses a large product database API for recognizing input and notifies users when their products are about to expire.

None of the applications discussed so far seem to aim at the behavior of users and only assist them while per-forming the tasks. Two application that do focus on behavior are presented by Ganglbauer et al. (2012) and Thieme et al. (2012). The former is an application that people can use to keep an electronic diary of their waste to help them reflect on their own behavior. Users can enter what they waste by taking a picture of it, select what reason it was thrown away for, and give a price of what was wasted. Users can see pictures and prices of what they wasted so far and find out what reasons they gave most. The main limitation is that the system was not tested yet, Ganglbauer et al. (2012) focused on de-signing the system using implications from interviews. Also, this system does not give a lot more than a list of pictures and reasons. It does not process the data or give more insights on the long run.

Thieme et al. (2012) discuss another, more general sys-tem. It aims to change behavior in food waste but also in recycling food. A mobile phone is mounted on the inside of the lid of a bin. The camera of the phone is activated by movement, so that it takes a picture of the bin content every time after it is closed. This fully au-tomates the “diary” functionality. Thieme et al. (2012) use crowd sourcing (Amazon’s Mechanical Turk) to an-alyze the pictures and extract data from them. This data is visualized as a tree with a varying number of green leaves and a varying number of gold bars. The number of green leaves represents how green the house-hold is by recycling waste and the gold bars represent how much they saved by not wasting food.

2.4

Persuasion & Behavior Change

Persuasion is defined by Fogg (2002) as an attempt to change behaviors and/or attitudes. Such a change in behavior is usually not something that happens overnight; it is a process that passes different stages. These stages are described by Prochaska and DiClemente (1982) in the context of people trying to stop smoking:

• thinking about changing (Contemplation) • becoming determined to change (Determination) • actively modifying habits/behavior (Action) • maintaining new habits/behavior (Maintenance)

Theoretically one would go trough these phases lin-early when changing behavior, but in practice one can progress, regress or stall in or between any of these phases. An important note on this paper by Prochaska and DiClemente (1982) is that it is written from a psy-chology perspective using therapy, whereas this research has a different approach using technology to change peo-ple’s behavior. However, they can be useful guidelines for the design of the system.

2.4.1

Captology

The overlap between persuasion and computers is called persuasive technology and the study on this technology is called captology (Fogg, 2002). Captology is described as a research field about how people are motivated and persuaded by the interaction with computer systems designed to changed people’s behavior and attitudes. Fogg (2002) explains that a computer can persuade in different roles, each containing its own subtypes. These roles are basically strategies that can be used to realize the planned behavior and/or attitude change. Some of these strategies will be used in the design of the system in the next chapter. The roles that a computer system can take to persuade are listed below.

• Role of a tool that creates capability. • Role of a medium that provides experience. • Role of social actor that creates a relationship.

2.4.2

Eco-Feedback

Persuasive technology has a lot of application domains, one of them being environmental conservation. Froehlich et al. (2010) call this field of research Eco-Feedback, which is an extension of persuasive technology. It is based on the based on the assumption that people are not aware of how their behavior affects the environment (Froehlich et al., 2010).

Froehlich et al. (2010) describes some strategies that have been used to motivate people for pro-environmental behavior, such as showing information, goal-setting, com-parison, incentives and feedback. This clearly shows that applications in the environmental conversation do-main are, more than others, focused on showing users how they behave. In fitness applications for example, the behavior (not moving enough), is already visible and the focus should be more on motivating people.

3.

PROBLEM STATEMENT

The related work section revealed some of the main is-sues and causes of food waste. All of these seem to find their roots in the fact that most people do not see food waste as a problem, which is no surprise because every-thing people throw away disappears in the bin. Most of

the widely available applications are focused on assist-ing users with specific tasks instead of changassist-ing their attitude and/or behavior.

The stages of behavior change, as described by Prochaska and DiClemente (1982), show that behavior change starts in the “Contemplation” stage, where people think about changing. It is quite obvious that people will not start thinking about changing their behavior when there is no reason or problem to do so. That is why showing the problem will be the main focus and starting point of the application. It will be like a diary for users where they enter their food waste.

Along with the diary function, some strategies to change behavior (by Fogg (2002)) will be implemented. These strategies contribute to showing the problem to the users, but also combine reflection and motivation to change their behavior. This should hopefully get users to the second and third stages of behavior change (“Determi-nation & Action”).

Concluding, the application will allow for reflection and motivate people to change their attitudes and/or behav-ior. The research questions therefore is as follows: Can a system that gives people insight in their own food waste change their attitudes and/or behavior towards food waste?

4.

SYSTEM DESIGN

The application will run on a tablet so that users can keep it within reach, probably in or around their kitchen. Users will be asked to register all their avoidable food waste. As this may be an annoying task that takes some time, the interface should be very simple and fast. Therefore the users will be allowed to guess the weight of wasted food. On the one hand this may be a limita-tion to the system, since the input can be faulty some-times. On the other hand it may stimulate people to think about how much they throw away.

Before users can input their data they need to be logged in. This allows all input to be clearly separated and the visualizations to be focused on their data only.

Registering food waste can be done at the “input” screen. It consists of 12 icons representing the different food categories used by Van Westerhoven and Steenhuizen (2010). When these icons are clicked a pop-up appears with a slider control to choose the weight of the wasted food and a drop down control to choose the reason. The user then presses “OK” and the system will store the data. The pop-up is closed and the system gives feed-back to confirm that the data was entered successfully. The input screen is shown in figure 1.

In addition to the diary functionality, the system has three main strategies to achieve the change in attitude

Figure 1: The input screen with icons for every food category.

and/or behavior. They are grouped by the different roles/strategies to persuade people, described by Fogg (2002). All three of them are subtypes of the “tool” and “medium” and include “self-monitoring”, “surveil-lance” and “cause-and-effect simulations”. Additional details about how these strategies work, how they have been used and what kind of effect they are expected to have will be described in the following subsections.

4.1

Tool: Self-Monitoring

In this role the system is a tool that allows users to monitor themselves. Self-monitoring makes it easier for users to see whether they are performing the desired behavior and helps them to understand their own be-havior (Fogg, 2002). In the context of food waste this role is very important, since everything we throw away disappears in the bin and is never seen again. This makes it very difficult to estimate how much we throw away and therefore how serious the problem is.

Tools for self-monitoring have been widely applied in an effort to reduce energy consumption and many dif-ferent types of reflection have been identified (Darby et al., 2006). Like eco-feedback, they aim at showing the consequences of the behavior to the user. This con-cept has shown to be very successful in reducing energy consumption, resulting in energy savings of around 5-15% (Darby et al., 2006).

The system allows for self-monitoring in different ways. The application has different views where users can see details about their own waste. All these views consist of bar chart visualizations. An example of a view can be found in figure 2. The different views are the following:

• Waste per day. Every time when a users registers waste a time stamp will be given to it. Therefore the system can show how much food is wasted per day. This allows the user to see the food wasted over time and maybe whether progression was made.

• Waste per category. Food is categorized, so all waste within the same category can be summed

up. This allows the user to see how much was wasted in every category.

• Waste per reason. All food waste thrown away for the same reason is summed up. This allows the user to see why food was wasted.

• Compare today’s waste in different categories with the average of in the Netherlands based on waste numbers by Van Westerhoven and Steenhuizen (2010).

Figure 2: Bar charts showing waste per day and per category.

4.2

Tool: Surveillance

Surveillance, in the terminology of Fogg (2002), is an-other role that the system takes. It is based on the fact that people behave differently when their behavior is being watched and it is a common and working way to persuade people (Fogg, 2002). To allow for surveil-lance, the system has a screen where the user can see the waste of the other households. Knowing that other people can see how much you waste should motivate them to show responsible behavior and give a feeling of guilt. The BinCam system also provoked this feeling of guilt, partly because their waste could be seen by other households (Comber et al., 2013).

Apart from the effect of surveillance, there is also a com-petition aspect in this strategy. Fogg (2002) mentions this may actually be the strongest type of intrinsic mo-tivation on a group level, which can motivate and ener-gize people. Both surveillance and competition had an effect on some of the participants in the BinCam study by Thieme et al. (2012).

One could argue that showing food waste of other house-holds is also Self-Monitoring, since it allows people to compare their own waste to that of other households to get a better view on their own waste behavior.

4.3

Medium: Cause-and-Effect Simulations

insight in the consequences of their attitude and/or havior immediately Fogg (2002). Showing the link tween cause and consequence can be very powerful, cause without the simulation the consequences of be-havior are often not visible. This is also the case for food waste. People know they throw something in the bin every now and then, but most of them probably do not know the exact consequences of this in amounts and costs on a yearly basis.

An example of a system using cause-and-effect simula-tions is Powerhouse (Bang et al., 2006). In this game users can see how different activities in the house (f.e. a long hot shower) affect the energy consumption, which should help them see the consequences of their behavior. This type of simulation is very complex and extensive. The system in this research will use a much more sim-ple type of simulation that does not allow for a lot of interaction.

The system has a screen with a very simple cause-and-effect simulation. The system calculates the expected weight and cost after a year based on the food that was wasted today. This can help people to see what the long-term consequences are of wasting small por-tions of food on a daily basis. A period of one year was chosen simply because yearly numbers are higher and therefore more impressing than smaller time peri-ods like weeks or months. The calculation of costs is based on the estimated costs of annual waste per per-son by Van Westerhoven and Steenhuizen (2010). This calculation does not make a distinction between differ-ent food categories, simply because the average costs of these food categories is not available. Every gram of food waste is therefore assumed to cost the same. This should not be a problem for the effect because people have the tendency to accept such a simulation as true and accurate (Fogg, 2002). The result of the simula-tion is shown in plain text, with a header saying “If you continue like today:”, as can be seen in figure 3.

Figure 3: Screen with simulation for yearly amounts/cost and comparing to average on bot-tom.

4.4

Technical Design

The system is built as a web application that runs full screen on a tablet. The application is fully written in

javascript, both on the server and client side. On the server side the application uses node.js and MongoDB (document database) to function as a web server and API. This serves all the files and data necessary for running the application on the client side. On the client side the application uses AngularJS, a javascript MVC framework that is maintained by Google. Most of the user interface elements are taken from Bootstrap with-out changing too much of the styling. The visualizations are powered by d3.js and nvd3.

5.

EXPERIMENT SETUP

To answer the research question the system was tested by 5 households of different sizes and compositions (see section 5.1). All of them were given a tablet to use during the 2 week (15 days) experiment. The system was installed on these tablets before they were given to the participants so that they were ready to use. Par-ticipants were also given two papers with instructions and some indications of what different foods weigh. Ex-amples include the average weight of an apple, average weight of a portion of pasta and the average weight of a full meal. This was meant to help people guess the weight of what they throw away when they save it in the system.

5.1

Households

All of the participants were acquaintances. Their house-holds have different compounds and are described be-low.

• Household 1 (hh1): 1 person. Working woman, mid twenties.

• Household 2 (hh2): 1 person. Studying woman, early twenties.

• Household 3 (hh3): 2 persons. Working couple, early fifties.

• Household 4 (hh4): 3 persons. Working couple with a son, early fifties. Son is almost 20 years old.

• Household 5 (hh5): 5 persons. Working couple with two sons and a daughter, early fifties. Chil-dren are in the range of 15-20 years.

5.2

Data Gathering

The data for this research was gathered in three dif-ferent ways. Two surveys (before and after the experi-ment) were filled in by all the people in the households to see how their attitudes, behaviors and insight in the problem changed because of using the application. The participants were asked to do the following in the ques-tionnaire.

• Reflect on how much they waste in 6 categories (opened products, after cooking, etc.). (questions used from Lyndhurst (2007))

• Rate themselves on 7 different “food skills” (shop-ping planning, cooking skills, etc.). (questions used from Lyndhurst (2007))

• Estimate (in numbers) how much their household wastes per year, how much it costs, and how con-fident they were about their estimates.

• Rate 7 statements about their behavior on a 5-point likert scale. This was a combination of pos-itive and negative statements such as “We always finish leftovers” and “I throw away food immedi-ately when it is passed date”.

• Answer 5 questions about how they feel about food waste (does it make them feel bad? do they put ef-for in preventing it? and why or why not?). (ques-tions used from Lyndhurst (2007))

Next to this questionnaire, an interview was done with all the people who really used the system. For example, in the 5 persons household, nobody used the system except for the mother. Therefore only she was inter-viewed. The interviews were used to evaluate the ap-plication and see what it was that made people change their attitudes/behavior (if they changed) or why it did not. This helped to evaluate how the strategies worked and whether they made people more aware and moti-vated to change their behavior.

At last, the data entered by the households was stored in a database during the experiment. After the two weeks of the experiment, this data was used to analyze the amounts of food wasted by the participants. It in-cludes all the reasons, amounts, timestamps etc. that were saved in the system.

5.3

Data Processing

Due to the low number of participants and some of the data being qualitative, the data is not processed in any way to show a significant change in attitudes or behav-ior. Instead, it is used to indicate if and how it changed people during the experiment. Proving the concept of this system would require a larger experiment with more households.

6.

RESULTS & EVALUATION

6.1

Data entered by households.

The data shows that the households throw away signif-icantly less than average. Taking the average of 43,7 kg avoidable waste per person per year (Van Westerhoven and Steenhuizen, 2010), the total waste of all house-holds combined (12 persons) would have been 21,5 kg in the 15 test days. During the experiment all house-holds combined registered only 13 kg. This could mean

that the system worked and really helped in decreas-ing the amount of food wasted. There are, however, other potential explanations for this. It is possible that the small selection of households for this study were already more active in preventing food waste than av-erage. Also, it is possible that participants estimated the weight of wasted food too low every time when they entered something into the system.

The most important reasons for throwing away food were “buying too much” and “not finishing leftover”, each accounting for over 3 kg of the total waste. The four categories accounting for the largest part of the total waste were vegetables, fruits, starches and diary products, which corresponds to the results by Van West-erhoven and Steenhuizen (2010).

Figure 4: Food wasted per day during the ex-periment.

A notable finding is that in the first week the waste is more evenly distributed than in the second week (see figure 4). In the first seven days all of the waste was above 600 g per day. In the last eight days, apart from peaks on three days, all were below 550 g. Even very low values like 250 g and 270 g were found here. The total waste in the second week was also slightly lower than the first week. This could mean people started wasting less after using the system for a while, but the differences are small and it is also possible that people became less consistent with entering waste in the system after a while.

The peaks are interesting though, since they often occur outside the daily routine. I know that hh2 wasted a lot in the end of the first week because she celebrated her birthday and hh4 wasted a kilogram of apples near the end of the two week experiment. I had the impression that the participants often felt excused for these peaks, almost like the waste was unavoidable. This is some-thing the system could focus on. It could, for example, separate these peaks from the regular waste and show them in a different screen to reflect on them.

6.2

Interviews

The six users who were (partly) responsible for entering the food into the system were interviewed. In every household this turned out to be one person, except in hh4, where there were two. All participants that did not really use the system will be referred to as non-users from now on. The aim of the interviews was to reveal possible issues that people had while using the system and to see how people reacted to the different

strategies to change their behavior.

6.2.1

Review of the System

None of the users had any technical issues with the sys-tem. Five of six specifically mentioned they found the system very easy and natural to use.

Five of six users admitted that they sometimes wrote down or remembered some of their waste to enter it into the system at a later time. For one of them (hh2) this was caused by problems with the wifi, which made it im-possible to open the applications at times. Apparently there were still situations where people found it easier to write down their waste instead of entering it into the system straight away. This could mean that they did not have the tablet within reach or that they found pen and paper easier than the application at times.

All users mentioned they had problems with the prede-fined reasons they had to choose from. These reasons were summarized from several researches about food waste behavior, discussed in earlier sections of this re-port. Apparently, they did not cover every possible rea-son and sometimes overlapped, which made it difficult for users to pick the right one. To solve this problem some reasons could be added to the system so that they cover most situations. However, situations where rea-sons are slightly different than usual will always occur and that is just a limitation to be accepted when using predefined reasons (classification). A more complex so-lution is that an extra option is added where users can type their own reasons. A potential problem, however, is that users could then add 20 different reasons. This would make it very difficult to compare the amounts wasted for every reason in a visualization.

One of the users (hh2) mentioned that she found it dif-ficult to guess the right amounts, especially in the be-ginning of the experiment. User 1 in hh4 said that he sometimes thought that other households guessed their amounts too low, since hh4 was leading in food waste. To solve this problem it is possible to weigh all the input. But then again, as with the classification of rea-sons, this would be a trade-off between ease of use and detailed information since weighing every input would probably take more time.

6.2.2

Behavior Change

All of the six users said that the system made them more aware of food waste. They had the impression that during the experiment they were wasting less food than they normally would during the experiment. The users adjusted their behavior in different ways and to different levels of fanaticism, but overall it was clear that they put more effort in simple things that pre-vent or reduce food waste. For example, hh1 stated she would consider eating something that is about to

pass date instead of throwing it a way. Hh3, on the other hand, had a full list of things she started doing differently because of the system. These, and changed mentioned by other users, were:

• Clearly buying less food than before, fridge was more empty (hh3, hh4, hh5).

• Putting more effort into saving leftovers (hh2, hh3). • Consuming food that they would have wasted

be-fore the experiment (hh1, hh3).

• Paying more attention to the planning of shopping, checking what is already in house (hh4).

• Shopping in smaller quantities and less tempted to buy discounted products (hh3).

During the interview the different strategies embedded in the system were explained. The users were asked to reflect on these different strategy and decide which one it was that changed their behavior. All of the users chose the comparison with other households. In con-trast to the other strategies, it was described as mo-tivating. Hh2 said she felt pressure because she knew that other people could see what she wasted, but it was the competition aspect of it that clearly motivated everyone the most. As participant 1 in hh4 said: “ev-eryone wants to win a competition”. Some users said they had contact with other users about the applica-tion. Hh3 and hh5 had sent each other a text message saying the “competition was a close call”. Hh3 men-tioned that she found this idea of “doing it together” attractive, because it allowed for interaction between her and other households. This suggests that adding a social aspect to the system, such as messaging each other or commenting, would be a useful addition for some users. Implementing existing social media is also an option, which was also done by Thieme et al. (2012) in their BinCam project.

The possibility to get more detailed information about personal food waste was the second strategy that worked well. Instead of being directly motivating, it was de-scribed as insightful and informative. This does not mean it is not motivating at all, hh5 said that seeing the actual numbers made the problem visible. It made her realize that she wastes a lot more than she thought. The “Simulation” strategy did not have a really good effect. Hh2 admitted she did not use this screen at all. User 1 in hh4 mentioned he could not compare the yearly amount and costs to anything, which made it kind of useless for him. Participant 2 in hh4 said she had seen it sometimes but she forgot what numbers were shown. This strategy could be improved by making the simulation more meaningful and interactive.

Another interesting finding is that 2 of 6 users men-tioned that entering the data into the system itself had an effect too. It made them conscious of the fact that they are throwing away food, instead of it being a simple action of throwing something in the bin. Hh3 said she felt like she had to “report” everything to the system when she wasted food.

I also spoke to some of the participants (both users and non-users) during the experiment (near the end). Many of them made jokes about finishing the plate and sparing waste until after the experiment, simply so that they did not have to enter it in the system. Hh2 told me that her mother (hh4) had eaten thawed ice redun-dantly simply because she did not want to throw it away. Apparently some people go quite far in chang-ing their behavior, almost leadchang-ing to behavior (eatchang-ing too much) that has other bad consequences.

6.3

Questionnaire

Since the questionnaire is only filled in by 12 people (twice), I am not going to look for statistically signifi-cant differences between the pre- and post-experiment questionnaire. Instead, I will only explore the data for notable findings that indicate that attitudes, behavior and/or awareness have changed during the two week experiment.

Participants were asked to guess how much their hold wastes per year, how much it costs their house-hold and also how confident they were that their guesses were right. Both before and after, only a few answered that they were “a little” confident about their guess, but others admitted they had no idea. Remarkable, however, is how close the average of their guesses was to the estimate by (Van Westerhoven and Steenhuizen, 2010). The average of guessed kilos of waste per person was 43 kg and 40,3 kg before and after the experiment respectively, which is surprisingly close to the 43,7 kg estimated by Van Westerhoven and Steenhuizen (2010). Apparently, the average estimated amount slightly de-creased during the experiment. However, when we split the data between users and non-users, we can see that the estimates by the users increased from 36,9 kg to 44,4 kg during the experiment and for non-users decreased from 49,1 kg to 36,3 kg.

The higher estimates by system users could be explained by the fact that they are the only ones being confronted with the actual numbers of the food waste, which makes them think their waste is higher than they thought be-fore. The non-users on the other hand, only see that their household is more active in trying to prevent waste than before, which makes their guesses lower.

When analyzing and optimizing the likert scale, the highest possible Cronbach’s Alfa when combining

state-ments was around 0,45, which cannot be considered re-liable. Even though this could be caused by the small sample size, this is not acceptable. When looking at the scores for the individual statements instead, minor changes can be seen. Slightly bigger differences can be seen when only looking at the data from the system users. Statements with a change of 0,50 or higher were:

• “I throw food away immediately when it is passed date.”, before: 3,33 after: 2,83.

• “If I do not feel like eating something that is about to go passed date, I throw it a way.”, before: 4.00, after: 2,83.

• “Leftovers are always finished”, before: 2,17, after: 2,67.

These numbers are only descriptive and nowhere near significant. They do, however, confirm some of the changes that participants mentioned in the interviews and they are clear examples of the simple things that people can do in an effort to prevent food waste. Most other questions in the questionnaire reveal no dif-ferences that are worth mentioning. They seem to ran-domly vary when comparing pre- and post-experiment results, but they roughly show the same answers. The only question where a small positive difference was seen is: “How much effort does your household put into min-imizing food waste?”. Responses before: a fair amount: 2, a little: 7, not so much: 3. Responses after: a fair amount: 4, a little: 8. This corresponds with the results of the interviews.

7.

CONCLUSION

It is safe to say that the system has changed the behav-iors of the system users. All of them mentioned they were more aware of food waste and (in different levels of fanaticism) changed their behavior in an effort to waste less food.

The design of the system was successful, since none of the users had remarks on the interface of the applica-tion. However, five of six users mentioned that some of the food waste was written down first before it was en-tered in the system. This suggest they sometimes found pen and paper easier than grabbing the tablet and using the application. This is remarkable, since apparently a tablet that is only used for the purpose of this research, and could always be left in the kitchen, was still not mobile enough.

The interviews revealed that surveillance and competi-tion were the best strategies to change the behavior of the participants. Comparing the total amount of food wasted with other households made participant compet-itive and motivated them to reduce the amount of food

they waste. The self-monitoring, including the entering of food waste into the system itself, was the second best strategy. It was described as informative and allowed users to reflect on their own behavior. The effect of the simple cause-and-effect simulation was disappointing, since participants mentioned it was a bit meaningless to them and some participants ignored it completely. The behavior changes were seen in adjustments that people made to their shopping, planning and food prepar-ing behavior. Some examples include buyprepar-ing less food, finishing leftovers or product that are past expiration data and paying more attention to planning. Appar-ently users all adjust their behaviors in different ways, so ideally the application should adapt its strategies to the user.

The effect of the application on the non-users was less apparent. The questionnaires did not reveal changes that are worth mentioning, except for the estimates of food wasted on a yearly basis. The questionnaire re-vealed that non-users estimated the waste 12.8 kg lower after the experiment, possibly because they had the im-pression that their household was paying more attention to reducing and preventing food waste.

8.

FUTURE WORK

Since this research is sort of a pilot study, there is a lot of room for improvements and extensions. It can function as a starting point for future work in different directions, some of which will be discussed in this section.

The results of this research are mostly indications. Only five households with twelve participants were tested dur-ing a relatively short timespan. This makes it a very useful study to get some first results on this relatively new research field, but more participants will be needed to get more reliable and representative results.

However, even with a much higher number of partici-pants, with the current experiment design we only fo-cus on attitudes and behavior by asking people about their behavior. In the end we would like to measure whether this behavior change actually leads to a lower amounts of food waste. The most objective way to test this would be to collect all the waste of the participants, as done by Van Westerhoven and Steenhuizen (2010) and WRAP (2008). The experiment would then also be split in three phases (before, during and after using the system) to be able to compare whether difference in amounts of waste can be found.

In addition, this new approach with waste collection and a higher number of participants would also allow to test the different strategies (surveillance, self-monitoring etc.) separately and compare them to see which one works best.

To simplify the input and processing of data, in this study a lot of classification was used. However, this may also limit the information that users would like to get from such a system. Further research would be needed to make a balanced compromise between the two; enough detail on the one hand, ease-of-use on the other hand.

The application can be be a small part of a full food managing system. Many applications to assist people while shopping, managing what they have in the fridge and finding recipes are already around. If these would be combined into a “food platform”, including reflection on food waste, they could potentially have a big influ-ence on the full food managing process. The waste that people enter in the system could play a role in making the system adapt its features and strategies to the user. For example, the system could offer help for shopping planning when a user is wasting a lot of food because he buys too much. The opportunities of social media in such a platform could also be explored.

Another point that would be of great interest for fu-ture research is combining such applications (either food waste only or the food platform) with smart fridges. Samsung already started selling smart fridges with a screen, WiFi and Apps, which means they could also run the application. This would be useful, since people would no longer have to keep a tablet in their kitchen. More potential, however, lies in building sensors into such fridges to actually know what products are inside it. That would take food managing to a completely different level.

Acknowledgments

First of all, I would like to thank my supervisor Frank for his enthusiasm and for providing me ongoing feed-back during this research project.

Secondly, I would like to thank Trifork B.V., in partic-ular Jason Fagan, for functioning as my personal help-desk while developing the application. Trifork has also provided me with the server to host the application dur-ing the two week experiment.

At last, I would like to thank Robbert-Jan Beun from Utrecht University for being the second reviewer of my thesis.

References

Bang, M., Torstensson, C., and Katzeff, C. (2006). The powerhhouse: A persuasive computer game designed to raise awareness of domestic energy consumption. In Persuasive technology, pages 123–132. Springer. Comber, R., Thieme, A., Rafiev, A., Taylor, N.,

Kr¨amer, N., and Olivier, P. (2013). Bincam: Designing for engagement with facebook for

be-havior change. In Human-Computer Interaction– INTERACT 2013, pages 99–115. Springer.

Darby, S. et al. (2006). The effectiveness of feedback on energy consumption. A Review for DEFRA of the Literature on Metering, Billing and direct Displays, 486:2006.

Fogg, B. J. (2002). Persuasive technology: using com-puters to change what we think and do. Ubiquity, 2002(December):5.

Froehlich, J., Findlater, L., and Landay, J. (2010). The design of eco-feedback technology. In Proceedings of the SIGCHI Conference on Human Factors in Com-puting Systems, pages 1999–2008. ACM.

Ganglbauer, E., Fitzpatrick, G., and Molzer, G. (2012). Creating visibility: understanding the design space for food waste. In Proceedings of the 11th Interna-tional Conference on Mobile and Ubiquitous Multi-media, page 1. ACM.

Godfray, H. C. J., Beddington, J. R., Crute, I. R., Had-dad, L., Lawrence, D., Muir, J. F., Pretty, J., Robin-son, S., Thomas, S. M., and Toulmin, C. (2010). Food security: the challenge of feeding 9 billion people. sci-ence, 327(5967):812–818.

Gunders, D. (2012). Wasted: How america is losing up to 40 percent of its food from farm to fork to landfill. Natural Resources Defense Council. Issues–Food and Agriculture. http://www. nrdc. org/food/files/wasted-food-IP. pdf (Page consult´ee le 6 janvier 2013). Lyndhurst, B. (2007). Food behaviour consumer

re-search: Quantitative phase. Technical report, ISBN 1-84405-383-0.

Prochaska, J. O. and DiClemente, C. C. (1982). Trans-theoretical therapy: Toward a more integrative model of change. Psychotherapy: Theory, Research & Prac-tice, 19(3):276.

Rouillard, J. (2012). The pervasive fridge. a smart com-puter system against uneaten food loss. In ICONS 2012, The Seventh International Conference on Sys-tems, pages 135–140.

Thieme, A., Comber, R., Miebach, J., Weeden, J., Kraemer, N., Lawson, S., and Olivier, P. (2012). We’ve bin watching you: designing for reflection and social persuasion to promote sustainable lifestyles. In Proceedings of the SIGCHI Conference on Hu-man Factors in Computing Systems, pages 2337– 2346. ACM.

Van Westerhoven, S. and Steenhuizen, F. (2010). Bepal-ing voedselverliezen bij huishoudens en bedrijfscater-ing in nederland. Amsterdam: CREM.

WRAP (2008). The food we waste. Technical report, ISBN 1-84405-383-0.