Final version: 20-8-2014, Supervisor: Frank Nack
Supervisor / first examiner: Frank Nack Signature:
Second examiner: Hannes Mühleisen Signature:
Motivating users to fill out
a food waste diary by using triggers
Ferdinand Richter, 5833310Thesis Master Information Science – Human Centered Multimedia University of Amsterdam Faculty of Science
Motivating users to fill out
a food waste diary using triggers
Ferdinand Richter University of Amsterdam f.p.j.richter@student.uva.nl
25 August 2014
ABSTRACT
Food wasted by individuals and households is a problem with a big environmental impact. Criteria for evaluating systems that support people in keeping track of food waste are established and used to examine four existing systems. Improvements derived from the evaluation are applied to a newly designed and built prototype. The new system employs several types of triggers in the form of notifications to motivate users to fill out the diary. A usability test is employed to test the new system. The test shows sending triggers makes it more likely participants will fill out their food waste diary.
Categories and Subject Descriptors
H.5 [Information interfaces and presentations]: User interfaces — Graphical user interfaces; H5.1 [Multimedia
Information Systems]: Evaluation/Methodology; H.5.2 [Information interfaces and presentations]: User interfaces — Prototyping.
General Terms
Measurement, Design, Experimentation, Human Factors.
Keywords
Food waste, interface evaluation, usability testing, electronic diary, behaviour modelling.
1. INTRODUCTION
Food wasted by individuals and households is a problem with a big environmental impact. Consumers could be supported by a system to become more aware about the problem and how they can contribute to reduce the amount of food wasted. This, in turn, makes the complete food chain more sustainable. The complete food chain, from production to consumption, constitutes a big part of the greenhouse gas emissions and raw materials usage. 17% of the greenhouse gas emissions can be contributed to food production and consumption in the EU and uses 28% of the raw materials delved [1]. The EU estimates that 40% of all the food produced globally is wasted. Preventing food waste means that more people can be fed with the same amount of produced food and a lower impact on the environment [2].
Where in the food chain the waste occurs and its impact on the environment differs between industrialized and developing countries. 40% of the food waste happens at retail and consumer level for the industrialized countries, while the same proportion of food waste happens after harvests and during food processing in developing countries. It is important to note that the same amount of food wasted later in the food chain has
a bigger environmental impact, compared to food wasted early in the process [2]. This thesis addresses food wasted by consumers, at the end of the food chain.
A study performed in 2013 suggests that the average Dutch household purchases 368kg of food per person per year. 19kg (about 5%) of this food is inevitably lost while preparing meals. Of the remaining 349kg, 47kg (about 13%) is wasted, while it would have been suitable for consumption [3]. The European Commission has given a working group1 the task to investigate how food losses can be reduced [4]. The working group cites the following reasons why food is wasted by consumers:
Lack of shopping planning.
Confusion about “best before” and “use by” date labels. Lack of knowledge on how to cook with leftovers. Inadequate storage.
Lack of awareness.
This thesis addresses the last reason cited by the working group: the general public has no knowledge of the food waste problem, its implications, how each individual plays a part in the problem and how they can contribute to reduce the amount of wasted food. By keeping a food waste diary people can become more aware about how much food waste they waste and how this is caused, thereby raising awareness in the process [5][6].
Currently, there are several systems that help users keep track of the food products they waste and offer some form of further insight. These systems are described in the section: “Related
work”, which also establishes the criteria used (a) to select the
best fitting system for a closer evaluation, performed in the section: “Food Waste Diary evaluation” and (b) are applied to the newly designed system in the section: “Design”. The new system is then tested, the section: “Evaluation” covers the methods used, and the results of, a usability test.
2. RELATED WORK
This section explores existing scientific literature on the main topics of this thesis. First, the general criteria for a food diary are established. Second, the Fogg Behavioural Model, a model on behaviour adaptation, is introduced. It describes the context that leads up to people performing a certain action. Third, four currently existing systems designed to help users keep track of food waste are described and then evaluated using the established criteria.
2.1 Food waste diary criteria
Food waste diaries can be used to gain awareness. This section addresses the criteria for such a diary in two parts: the design
1 Its complete title: “Working Group on Food Losses/Food
Waste of the Advisory Group on the Food Chain, Animal and Plant Health”.
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
conventions for an electronic diary in the context of food waste; second, which data needs to be gathered.
2.1.1 Electronic diary design
Two types of traditional diary design can be distinguished:
time-based and event-based. Time-based diaries are filled in
with a fixed time interval; event-based diaries are filled in when a specific event has occurred [7]. As food waste does not necessarily happen daily and can occur at any time an event-based approach is preferred.
Studies on electronic diary design reveal that participants preferred a system that is flexible: it should allow them to fill out the diary at a time of convenience and it should be possible to revisit and alter previously made diary entries [7].
2.1.2 Desired data
What data is needed and when it needs to be gathered from the diary depends on the research question, of which three general types are identified [7]: (1) What are the experiences from an individual, and how do those experiences vary in nature and over time? (2) How do experiences vary between people? (3) What is the cause for the differences between people? In the context of food waste, the main questions that need to be answered are: How much food do I throw away? How does my waste pattern compare to that of my peers? How does my food waste pattern change over time? In order to gain a picture of how much food is wasted by an individual, the following data has to be collected:
Date & time of wastage.
Type or category of the food product. The reason the food was disposed of. The mass of the discarded food. Geographical data.
Recording date & time of wastage allows keeping track of how the food waste pattern changes over time. The reason the food was discarded says something about whether or not it could have been avoided to throw away a certain food item. Research shows that there are differences in the food waste between people living in urban areas and those living outside them [3]. In order to make a fair comparison between peers (e.g.: comparing urbanites to fellow urbanites), geographical data can be used, eliminating the need to ask the user to fill out a questionnaire about personal background information.
2.2 Behaviour adaptation
The Fogg Behaviour Model (FBM) describes the prerequisites for a change of behaviour that need to exist before someone initiates an action [8]. It features two dimensions: motivation and ability, and a trigger. Motivation ranges from low to high and says something about how willing someone is to perform a certain action. Ability expresses to what level someone is able to perform an action, it ranges from: “the action is hard to perform”, to: “the action is easy to perform”. When someone is sufficiently motivated and has the ability to perform an action, it is then prompted by an appropriate trigger.
The core motivators consist of the following emotions and sensations: pleasure and pain, hope and fear, social
acceptance and rejection. Fogg notes that some of these
motivators – hope, for instance – are more practical and ethical to employ than others, such as: pain. As keeping track of food waste is something that will need to happen voluntarily, it would be unwise to motivate users with any of the negative motivators.
Ability is described by the “elements of simplicity”, which are:
time, money, physical effort, social deviance, brain cycles and
to what extent a certain action is non-routine. When the action requires too much of any of these elements the ability to perform the action the threshold is considered to be too high. The “scarcity” of the elements of simplicity differs per individual for the same action: one person might be willing to spend money, but not time or brain cycles, while for someone else, the opposite might be true. Logging food waste should be as streamlined as possible, requiring minimal time and effort and should feel as routinely as possible.
Three types of triggers can be distinguished: the spark, the
facilitator, and the signal, each applicable in different
situations. When a person has both adequate motivation and the ability to perform the action, a signal is sufficient to tell someone this is an appropriate time to perform an action. A
spark can be used to motivate someone when motivation is
low, by instilling any of the previously listed emotions and sensations. A facilitator is a trigger that somehow conveys the message that the target action is easy to perform and that does not require too much of the elements of simplicity. High and low motivation and ability are subjective measures, which require user modelling and can only be established by looking at people’s behaviour, or interviewing them. Motivation and ability can, however, still be influenced, until they at some point are high enough for someone to perform an action. Due to the ubiquitous presence of computers, smart phones and other media, people are nowadays confronted with so many triggers that they are increasingly considered to be a nuisance, rather than informative [8]. This has to be taken into account in the design of the system: only trigger the user when the motivation to perform an action is as high as possible. It is important to note that the behavioural change model can be used in two situations: as a way to support an action, or as a way to discourage performing an action. In the context of food waste prevention, this could mean that someone is more likely to perform the action of filling out a food waste diary, as applicable in this thesis, if it requires little effort, for instance. The opposite situation, someone might be dissuaded to buy excess food, by showing that it costs money and that it requires effort to purchase the food, transport it home and transporting it to the waste bin. In the context of this thesis, users are stimulated to perform the action of filling out a food waste diary.
2.3 Existing systems
Using the behaviour adaptation and designing for reflection frameworks four systems are reviewed.
2.3.1 BinCam
The BinCam system consists of a trash can (or “bin”) and an application on Facebook [9]. The bin has a smart phone attached to the bottom of the lid (seen in Figure 1), which automatically snaps a picture every time something is thrown away [10].
Figure 1. BinCam’s bin cam. The camera is attached to the bottom of the lid.
The picture is uploaded to Facebook immediately, where it is displayed on the photo stream of the user’s BinCam identity2
. The picture is also sent to Amazon’s Mechanical Turk, where it is analysed. The crowd source platform is used to extract the following from the picture: the total number of items, the number of recyclable items and the number of food items. This information is then applied to the picture as a tag, which users can edit. Users can also delete the picture.
Users receive scores calculated from crowd sourced tags. The scores are used to create the BinLeague (Figure 2), in which several users compete against each other. There are two competitions: “recycling achievements” and “preventing food waste”. The first is visualised by a tree, where less leaves means the user did a bad job at recycling. The second is visualised with gold bars, where a pile of gold bars means the user saved a lot of money by not wasting food.
Figure 2. BinCam’s BinLeague. Scores are shown by using trees and gold bars. More leaves on the trees means a lesser environmental impact, more gold bars means more of a
monetary reduction.
The system is flexible: users can choose to revisit and delete diary entries. However, entries cannot be edited. The system does not gather exact data on how much food is discarded, unlike other systems, nor does it ask for a reason. It is unclear whether or not the system used a method to compare the food waste pattern of the user to one of his or her peers. The system does, however, allow people to compare how they fare by using custom groups of people.
While, from a technical point of view, the system seems to function adequately, there are two factors that seem to hinder acceptance by the public of BinCam: storing and promptly displaying of all the pictures made can be perceived as too much of an intrusion in the users’ private life3
and the shaming aspect of the competition, which punishes users that are not performing as well as their peers4. BinCam is the only system,
of the four systems analysed, that takes into account the motivation dimension of the FBM. The way it is implemented, however, might actually deter people from using the system instead of motivating them to use it.
The question also remains whether or not the out-of-sight automated analysis by Mechanical Turk and posting of all the pictures on Facebook makes it easy to forget about BinCam,
2
The BinCam Facebook URL no longer works, it has either been taken off-line, or is experiencing ongoing technical difficulties: https://apps.facebook.com/thebincam/. It has unfortunately not been possible to examine in this thesis. All research is based on research papers.
3 http://www.theguardian.com/environment/blog/2011/jun/08/ camera-bin-waste-facebook-recycling 4 http://www.engadget.com/2011/06/10/bincam-posts-photos-of-your-trash-on-facebook-shames-you-into-r/
and whether or not the logged pictures are ever reviewed again. A hands-on evaluation could determine whether or not this is the case, such an evaluation is, however, not performed for this thesis.
2.3.2 LeanPath
The LeanPath system is aimed at large kitchens of restaurants and emphasises on correctly and individually logging the food. The full system consists of a scale and a touch interface, the mobile version only employs a tablet.5
The user puts the food product(s) to be disposed on the scale and the camera mounted above the scale snaps a picture. The user then provides the following information:
Employee name.
Food category & specific type of food. Reason of disposal.
Type of container (so its weight can be deducted).
Figure 3. A complete LeanPath system, with the touch screen on the left and the scales with camera attached on
the right.
The system provides an initial overview, displaying the data provided by the user, together with data generated by the system (weight, value and the picture). After submitting the report, the system shows how the monetary value of all the food submitted by the user that week; how much money can it would cost if that specific item was thrown away every day of one year; and the environmental impact, measured in “cars on the road”, gallons of water or gallons of oil. The system displays the cost of the products on the scale and presents the user in the next screen with a total of all products processed by the system today. Users can go back and change previous entries for a limited time if they have made a mistake. After submitting the report the system provides an overview of progress towards a specific goal (for instance: reducing soup waste). The system also shows a list of “top waste watchers”, users that submitted the most waste in weight.
Except for geographical information, all necessary data is collected. The integrated scale ensures that minimal effort is required to fill out a diary entry.
2.3.3 Wise up on Waste
The application (displayed in Figure 4) is aimed at professional kitchens to help monitor food waste, as part of a sponsored program6. The application also includes recipes and tips to reduce waste, such as how to store fruit and vegetables. 5 http://www.leanpath.com 6 http://www.unileverfoodsolutions.co.uk/our-services/your-kitchen/wise-waste-app
Users provide the following data for every breakfast, lunch and dinner:
Date (defaults to today’s date). Number of covers (plates ordered).
Three types of waste in kilograms: food lost due to
spoilage, during preparation, or left on the customer’s plate.
After using the application for a number of days, an audit
report can be generated. It states for every meal course: (a)
average waste per day, (b) average waste per plate, (c) a breakdown of the three types of waste and (d) a non-negotiable 20% waste reduction goal per week, expressed in kilograms and monetary value. The waste breakdown of several audit reports can be compared using the audit history, which also features the progress towards the 20% waste reduction goal. Where the diary design of all the other evaluated systems is event-based, this system only allows the user to report time-based. This means that users cannot fill out the diary whenever some food is discarded, but instead have to remember it or write it down for when they diary is filled out.
Figure 4. Wise up on Waste overview pages. The left image shows the data for one particular meal as provided by the user, summarized in a donut-chart diagram and provides a
food waste reduction goal. The right image shows a breakdown of multiple meals.
The system raises awareness concerning the performance compared to other audit periods, but the system does not ask for information in regards to which food products are being wasted specifically. This leaves users guessing as to how to improve food waste performance. Previously made entries cannot be revisited and edited.
The application takes considerable time to start (well over 20 seconds on an iPhone 3GS) and feels sluggish in use. After the application has loaded, a menu is displayed. The “Record” menu-item is selected, in order to make a diary entry. It takes about 10 seconds until the diary entry screen appears. This makes logging one diary entry more time consuming compared to the other systems. The user has to select a monetary value for all the food wasted. Trying to calculate the appropriate amount costs time and requires extra effort. As there is no standardized way of getting the cost, this part is considered
non-routine. This could lead to users estimating the value,
which increases the error rate of the results, rendering them less usable.
2.3.4 Food Waste Diary
Users can log their food wasting habits by using the Food
Waste Diary (Figure 5). Apart from the user’s geographical
location and the date, which are logged automatically, users provide the following:
The reason the food was thrown away from a predefined set: visibility of food stock is missing, over-buying, no
shopping list, change of cooking plans or special offer,
or users can select a custom reason.
A food category, again from a predefined list or by adding a custom category.
A picture of the food in question, the application opens the camera. Note that an earlier taken picture cannot be selected.
A price.
There is space for additional comments.
After having made one or more log entries, users can review their own food wasting data by looking at a (a) chronologically ordered list (newest entries first), (b) a gallery of pictures taken and (c) a bar chart which displays the occurrence of the reasons food was thrown away. The same three types of displays can also be used to review the combined data from all the users using the system. The most occurring reasons listed are: over date, over-buying and no shopping list. Logging the reason food was disposed of is hindered by the fact that over
date is not one of the default reasons included. Even though
the list of reasons users can expanded the list with custom reasons, the list itself is incomplete as long as over date (or
expired) is missing.
Figure 5. Food Waste Diary. The left image shows a new entry. The right image shows the picture gallery as it was
intended by the creators.
It must be noted that as the overview of all the users does not include any pictures later in the log, it can be concluded that either (a) users do not bother to take the pictures, or (b) this feature was subject to abuse and has since been disabled by the system’s creators. The first reason could be explained by the fact that it is not possible to select any pictures from the phone’s picture library. It might not be feasible to open the app, snap a picture, fill in and submit the report and do all these things repeatedly while, for instance, preparing dinner. The user is denied the choice of quickly snapping a picture and filling out the report at a later time of convenience. It is also not possible to edit a report afterwards in order to add data or a picture. An application designed to track which types of food
are wasted, has to at least give the user ample opportunity to do so. It is impossible to change or delete diary entries.
Figure 6. The picture gallery as it looks today.
The default food categories are not as descriptive compared to the selection options in the LeanPath system: where in
LeanPath, the exact food product can be selected, Food Waste Diary only allows selection of categories. Similar to the
reason, the food categories can be customized. However, this requires extra effort from the user every time a certain food product is logged for the first time. It follows from the FBM, that as actions require more effort, the chance they are performed diminishes.
After logging the food category the only place it appears is the chronologically ordered list. Without a picture or a specific food category, a user cannot tell which food products are, for instance, bought in excess and therefore wasted when they are not consumed. As the system features both a screen with personal and a screen with a user base wide overview of reasons, users can compare their food waste reasons with that of everyone else. However, it would be much easier if both overviews were integrated on the same page.
The list and pictures of all the users bundled together really do not convey any useful information, as they are just streams of rubbish.
2.3.5 Diaries evaluations summary
This section summarizes the evaluation of pre-existing systems. Table 1 shows an overview of how the four evaluated systems fulfil the practical criteria for a food waste diary.
Table 1. Diary criteria.
Event-based Time-based Flexibility Data
BinCam × * ○ -
LeanPath × ○ +
Wise up on Waste × - -
Food Waste Diary × - +
×: feature is present
-: insufficient; ○: sufficient; +: more than sufficient
As stated earlier, Wise up on Waste is the only system that forces its users to report time-based. Wise up on Waste and
Food Waste Diary do not offer any flexibility in reporting. LeanPath and BinCam perform better in this regard, but they
still do not offer the user complete control over previously made diary entries.
Tabel 2 displays how the systems compare using the criteria derived from the FBM. All the systems, except Wise up on
Waste, allow their users to make a diary entry using minimal
time and effort. None of these systems uses triggers in any form.
Table 2. Behavioural criteria.
Motivation Time Effort
Non-routine Trigger
BinCam - + + + -
LeanPath ○ + + + -
Wise up on Waste ○ - - ○ -
Food Waste Diary ○ + + - -
-: insufficient; ○: sufficient; +: more than sufficient
Of the four systems described, LeanPath is the best match to the criteria for a food waste diary. While the application allows its users to log all the information established in the food waste diary criteria, there are a few instances where it could be improved.
All systems are lacking a trigger in any shape or form. On the task of filling out the diary, a signal can be enough to remind users and, as the systems do not contain a motivating component, a spark can be used to motivate users.
A complete system with integrated scales might a bit big to use in a non-professional kitchen. However, scales in some form or shape can already be part of kitchens. It should be considered to integrate a way to input the precise weight of the discarded food into the newly designed system, by using regular kitchen scales.
3. DESIGN
In this section the design of the Food Waste Diary Prototype (FWDP) is presented. The design is based on the criteria as they are derived from studying the four existing systems, combined with general diary criteria and the FBM.
None of the examined systems contain a trigger. None of the systems attempts to motivate its users with the use of a spark (which would be the type of trigger sent if you want to motivate a user), nor do any of the systems send a signal to remind the users to perform the task. The goal of the FWDP is to test how users adhere to filling out the diary if they do receive these two types of triggers.
The questions the FWDP can answer are:
Are triggers effective in reminding and motivating users to fill out their food waste diary?
What is the effect of different types of triggers? The FWDP consists of a web site7, designed to be used on smart phones and a notification application. It is constructed using the Axure rapid prototyping platform8. The prototype functions on both the Android and iOS platforms, where it behaves like a regular smart phone application.
3.1 Interface
The landing screen of the prototype is displayed in Figure 7. This is the diary entry system and it allows users to input a food type, the reason the food was lost and its weight. When the user is done filling in the diary, Done is pressed, and a text is displayed, indicating that the entry has been saved.
7
http://nrichter.home.xs4all.nl/foodwastediary/
8
Figure 7. The New Diary Entry screen. Users can select a
Food Type, a Loss Reason, enter a Weight, select if the
weight is in grams or kilograms and press Done when they are ready filling out the diary entry.
Figure 8 displays the Food Type selection menu. Following the LeanPath example, users can select one of the following categories: bread, fruit, vegetable, dairy product or oil & fat. If any of these categories does not reflect the food being disposed of, users can select other.
Figure 9 displays the Loss Reason selection menu. The following reasons can be selected: expired, overproduction,
spoilage, trim waste and other.
As the purpose of the prototype is to evaluate triggers in the context of a FWDP, the prototype features the minimum requirements for a food waste diary. The diary is event-based, the other requirements: data and flexibility are present in a minimum way necessary to carry out the experiment.
Figure 8. Users select the type of food. After selecting any of the items, the system returns to the home screen, where
the Food Type is replaced with the users’ input.
It follows from the FBM that if users have to spend less time on a task, the bigger the chance that task is executed. This means the possibility to include a picture is not included in the diary prototype (data); neither is adding a location where the waste occurred (also data). Also, users cannot review and edit previous entries (flexibility). For the purpose of the
experiment, the interface has been translated to Dutch, as the experiment features Dutch participants only.
Figure 9. Just as with Food Type, after selecting the Loss
Reason the system returns to the home screen, where the Loss Reason is displayed.
3.2 Triggers
The content of the triggers follows from the FBM. There are four different triggers: one signal (1) and three different
sparks: pleasure (2), hope (3) and social (4). Translated from
Dutch, they read as follows:
1. Disposed of food? Fill out the food waste diary on: <link>.
2. By discarding less food, more money remains to spend on fun things! Fill out the food waste diary on: <link>. 3. Become more aware of how much food is discarded
and save money and the environment. Fill out the food waste diary on: <link>.
4. Do you discard more or less food than your peers? Find out! Fill out the food waste diary on: <link>. The first trigger is considered to be a signal, simply reminding the user to fill out the diary. The other triggers are sparks, of three different flavors, following the FBM, motivating the user in different ways. Trigger 2 is supposed to remind the user that filling out the diary brings pleasure, trigger 3 instills the hope that money and the environment can be saved and trigger 4 focuses on the social aspect of filling out the diary and discarding less food. Keeping the negative response to BinCam’s “shaming” strategy in mind, only positive triggers are employed.
Prowl9 is used to receive triggers on iOS, Notify My Android10 is used on Android. Users have to install these applications in order to receive triggers in the form of notifications. The notifications contain a text, stimulating or just reminding the user to fill out the FWDP, and a link to the diary itself. The notifications appear on smart phones the way regular notifications from text messages do.
Prowl and Notify My Android accounts are set up for each
individual user, prior to the experiment. Users have to log in to these accounts, so the experiment leader can (manually) send
9
http://prowlapp.com/
10
triggers at a by the user self-determined time, and the message can be different for each user.
3.3 Tracking user behaviour
A separate folder on the prototype website server is created for each individual user. This allows tracking users separately using Google Analytics (GA). All the interactive elements are tied to GA events, so every user action can be tracked.
4. EVALUATION
A user test is performed with 22 participants, 10 female, 12 male. Their ages vary between 20 and 64, with an average of 28.9. It can be noted that while two participants have an age of 42 and 64, the age of the others vary between 22 and 35. Excluding the two outliers, the average age moves to 26.6. Prior to the experiment, participants are invited to answer a survey. Participants were then asked to use the FWDP for five days duration. After the experiment, a structured interview was conducted with three participants.
4.1 Survey
The survey consists of 13 questions. Question 1 establishes if the participant has read the experiment documentation and the informed consent form. With question 2 participants can indicate at what time they wish to receive notifications. Question 3, 4 and 5 ask if the participant is the one responsible in the household for daily groceries, how often he or she shops for groceries and often it happened the last seven days. Question 6, 7, 8 and 9 ask if the participant is responsible for the disposal of food, how often (on average) throwing away food occurs, when it happened during the last seven days and how much he or she estimates was thrown away the last week. With question 10, participants are asked to estimate how much food is thrown away on a yearly basis in the average Dutch household, in order to reveal to what level the participant is aware of the food waste problem.
Question 11, 12 and 13 establish to what level participants are motivated to prevent food waste. These questions feature a five point Likert scale, where participants are asked of their opinion on: if they think food waste prevention is an important subject; if they want to actively prevent food waste; and if they think keeping a food waste diary is a good way to gain insight in how much food they waste.
4.2 Experiment
Participants are invited to use the FWDP and fill out the diary for five days. They were instructed to log the activity of throwing away food, indicating the type of food, the reason it was thrown away and the weight. The participants all start on the same day.
In the five days of the experiment, they receive one trigger every day on four days and no trigger on the remaining day. Receiving no trigger functions as a control, so the influence of triggers can be measured. The order in which the triggers are sent is randomized, as well as the content of the triggers, to eliminate order bias.
The trigger is sent on the time indicated by the participants in the survey. For the four participants that did not fill out the pre-experiment survey a default time of 18:30 has been selected. This time has been chosen around dinner time, leaving enough time to fill out the diary before bed time. Prior to the experiment, it became apparent that the FWDP did not run properly in iOS Safari, if the iPhone was not connected to a stable WiFi network. This was most likely caused by the
bulky JavaScript, included in the FWDP, which was set to load prior to the rest of the web site. It has been considered moving the JavaScript part to the end of the main HTML file, which caused the web site to load considerably faster, at least visually. It would, however, leave the user with a dysfunctional prototype, and prevent tracking any of the participants’ actions. This strategy has been abandoned and the experiment continued with the original prototype.
Loading times were considerable, taking up to and (depending on the quality of the WiFi connection) over one minute, which is considered unacceptable for participants to continue with the experiment. If a task requires more time, it follows from the FBM that there is a smaller chance that the task is performed. Initial instructions included disabling the WiFi connection, and using 3G instead. This did help in some cases, but still participants indicated long loading times.
To solve this problem, participants were instructed to install Chrome and run the FWDP in that browser, before they started the experiment. Loading times improved considerably, even if the WiFi connection was unstable.
4.3 Structured interview
After collecting and processing the experiment data, more data was needed to provide an explanation to the outcome of the experiment. Three weeks after the experiment has been concluded, three participates were approached for a structured interview, in order to uncover additional data.
The questions asked ranged from whether the notification system functioned in a way that compares to regular notifications, to their reaction to the content of the notifications, as well as more general questions on filling out the FWDP.
5. RESULTS
This section presents the results of the survey, the experiment and the structured interviews.
5.1 Survey
18 of the 22 experiment participants participated in the survey. 10 of the respondents are female, 8 are male.
On the question: “Are you responsible for the household
grocery shopping?” 7 participants respond they are, 5 respond
they are not and the remaining 6 respond that it is a shared responsibility.
On the question: “How often do you shop for groceries” participants respond an average of 5.06 times per week. 10 participants respond they are grocery shopping 6 times per week. The remaining 8 participants account for shopping in the range of 2 to 5 times per week. It must be noted that participants who respond they are responsible for the household groceries, also respond that they shop 6 times per week.
Table 3 shows the relation between how many times a week a respondent thinks he or she shops, compared to how many times they shopped the last 7 days: 1.94 on average. Although it seems respondents overestimate how many times a week he or she goes out shopping for groceries, the results are consistent: respondents that think they shop more times a week, also report having shopped more the last 7 days. To the question: “Are you in your household responsible for
the disposal of food?” 4 participants states they are, 1 is not,
and the remaining 13 states they share the responsibility. The 4 participants that state they are responsible for disposing food, also state they are responsible for doing the household grocery
shopping. Participants seem to overestimate how many times a week they shop for groceries, when compared with how many times they reported shopping in the last 7 days.
Table 3. Respondent reported grocery shopping, compared to past 7 day average.
n 1 1 4 2 10
Respondent estimation 2 3 4 5 6
Past 7 day average 1.0 1.0 1.5 1.5 2.4
n: number of participants
Table 4 displays how many times the respondents throw away food, compared with how many times they reported to have thrown away food the last 7 days.
Table 4. Reported estimate of number of times throwing away food, compared to the last 7 days.
n 7 day average
Less than once a month 1 1.0
Monthly 2 1.0 2-3 times a month 5 2.2 Weekly 7 1.6 2-3 times a week 2 2.0 Daily 1 3.0 n: number of participants
The results of these questions are less consistent as the ones regarding grocery shopping. This could be explained by the fact that shopping often leaves evidence, like a receipt, or the groceries itself, which can be reviewed at by the respondents as they fill out the survey. Disposing of food eliminates evidence, leaving nothing if the trash has been taken out, nor is it very appealing to go dumpster diving to remind oneself of what was thrown away, even if it concerns ones one trash can. This result does, however, enforce the idea that people are not aware of the food they throw away.
On the question how much food was thrown away in weight one participant reported throwing away 500g–1kg, 6 reported 250g–500g and the remaining 11 reported throwing away 250g or less. This seems to be on the low side, as research shows that the average household throws away 66kg of food each year (19kg trim waste + 47kg preventable waste), which is about 1,3kg per week (using 52 weeks per year). Household size was not included in the survey, but the average Dutch household consists of 2.2 people [11]. Even when correcting for a household size of one, which would mean a Dutch person on average would throw away 0.58kg of food, the figures reported seem to be on the low end. This could indicate that participants, indeed, are not aware of what they throw away. The next question tries to establish to what level the respondents are aware of the food waste problem, by asking to provide an estimate of how much food is annually wasted by the average Dutch person.
In order to uncover to what extent respondents were motivated to “do something” about the food waste problem, they were asked three questions, where answers could be given on a 5 point Likert scale, ranging from “strongly disagree” (1) to “strongly agree”. Figure 10 shows how the respondents responded.
On question 11, respondents on average responded with a value of 3.94, with a standard deviation (sd) of 0.87. On question 12, the average is 3.70, sd=1. On question 13, average is 4.00, sd=0.77. The overall average of the three questions combined is 3.90, with sd=0.7.
Overall, it can be concluded that respondents think food waste is an important issue (Q11), they want to do something about it (Q12) and, most of all, think that a keeping food waste diaries is a good way to do it (Q13).
Figure 10. Motivation responses.
5.2 Experiment
During the 5 days of the experiment, user interactions have been recorded using GA. 18 participants used the Android smart phones, 4 used iOS. Table 5 displays the cumulative FWDP visits across all participants.
Table 5. Cumulative FWDP visits for each type of trigger, as well as no trigger. Day 1 2 3 4 5 Total A (no trigger) 1 1 2 4 0 8 B (signal) 1 5 5 4 0 15 C (spark: pleasure) 4 3 2 0 2 11 D (spark: hope) 2 4 3 0 6 15 E (spark: social) 13 2 1 3 4 23 Total 21 15 13 11 12 72
Note that <link> has been replaced with the participants’ own URL, enabling tracking by GA. The 22 participants visited the FWDP a 72 times in total, 3.27 times per participant over the course of 5 days, 0.65 times per participant per day.
Following these results, it appears sending a trigger definitely draws attention to the FWDP. Where the control group (no trigger) visits the FWDP 8 times, the groups that did receive a trigger visit the FWDP 16 times on average. A clear drop-off near the end of the experiment is also visible: on the first day participants visit the FWDP 21 times, declining steeply to the second day with 15 times and slowly diminishing to 12 times on the last day of the experiment. The control group does not show this clear decline over time, indicating that the effect of the trigger wears off even as soon as three days after sending the first trigger.
Visiting the FWDP did not mean participants actually filled out the diary. Likely due to the prototype taking a long time to load, participants repeatedly refreshed the FWDP web page. Another possibility is that the FWDP was opened after a trigger was been received, as the trigger contained a link to the FWDP, which the participants could follow. In Table 6 and Table 7, these events have been cleaned out, showing only singular instances of participants visiting the web page.
0 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 Q11 Q12 Q13
Studying the events tracked by GA, many instances of visiting the FWDP had to be filtered out, resulting in Table 6. This leaves 47 instances of participants filling out the diary, 2.14 times per participant over the course of 5 days, 0.43 times per participant per day.
Table 6. Recorded instances of participants filling out the diary, for each type of trigger, as well as no trigger.
Day 1 2 3 4 5 Total A (no trigger) 1 1 2 3 0 7 B (signal) 1 4 3 3 0 11 C (spark: pleasure) 2 2 1 0 1 6 D (spark: hope) 2 3 2 0 3 10 E (spark: social) 6 1 1 2 3 13 Total 12 11 9 8 7 47
Although more ambiguous than the results from FWDP visits in Table 5, the event based counting in Table 6 still reveals triggers have an effect. The control group filled out the diary 7 times, where the groups that did receive triggers filled it out 10 times on average per trigger. The decline visible in Table 5 is also visible in Table 6, though far less steep. This can in part be explained by a sole participant, whom visited the FWDP 7 times within the same hour, but only filled out the diary once on the first day.
It can be debated what size interval between receiving the trigger and filling out the diary should be considered. It has been 24 hours thus far, but can a trigger really be viewed as such if it takes a participant nearly a day to respond? Table 7 displays the instances where the diary was filled in within the first 3 hours of receiving a trigger. The same time period was considered for the no-trigger control group, using the time they would have received a trigger if they were in any of the other groups.
Table 7. Filling out the diary 3 hours after trigger time.
Day 1 2 3 4 5 Total A (no trigger) 0 0 1 2 0 3 B (signal) 4 2 3 2 0 11 C (spark: pleasure) 2 1 1 0 1 5 D (spark: hope) 0 3 0 0 2 5 E (spark: social) 5 1 0 1 1 8 Total 11 7 5 5 4 32
A total of 32 instances of filling out the diary can be discerned, averaging 1.45 per participant over the course of 5 days, which equals 0.29 times per participant per day. The diary has been filled out 7.25 times per trigger over the course of 5 days, compared with 3 times when no trigger has been sent. The results are consistent with the results displayed in Table 5 and Table 6: sending a trigger leads to more adherences to filling out the diary and adherence tapers off towards the end of the experiment.
The results of Table 5, Table 6 and Table 7 are summarized and compared in Table 8. The results clearly show a positive relation between sending a (a) trigger and (b) visiting and filling out the diary, compared with receiving no trigger. This result most clearly surfaces when looking at the 3 hour interval after sending the trigger.
How the individual triggers compare is summarized in Table 9. To easily compare them, they have been normalized by dividing the number of events per trigger (or control group) by the total number of events.
Table 8. Comparing visits and events within 24 hours after sending a trigger with 3 hours after sending a trigger and
control groups.
visits (24h) events (24h) events (3h)
total events 72.00 47.00 32.00
average per participant 0.65 0.43 0.29
average per trigger 16.00 10.00 7.25
no trigger 8.00 7.00 3.00
trigger/no trigger ratio 2.00 1.43 2.42
Table 9. Normalized trigger comparison.
visits (24h) events (24h) events (3h)
A (no trigger) 0.11 0.15 0.09
B (signal) 0.21 0.23 0.34
C (spark: pleasure) 0.15 0.13 0.16
D (spark: hope) 0.21 0.21 0.16
E (spark: social) 0.32 0.28 0.25
When the triggers are compared, it appears that two triggers consistently score higher than the remaining two: the signal and the social spark perform better than the pleasure spark and
hope spark, by a ratio of 1.47:1 (visits 24h), 1.50:1 (events
24h) and 1.84:1 (events 3h). These ratios indicate that, on average, a signal and social spark are more effective than the
pleasure spark and hope spark.
There is not enough data to say something about an individual’s preferences, as each trigger was tested only once per participant.
5.3 Structured Interview
The interviews were conducted with one female participant and two male participants. One interviewee used iOS, the other two Android.
Participants indicated that they did receive the triggers and the liked the fact they could “customize” at what time these triggers should be received. They did report that in some instances on Android systems, the notifications did not stand out enough, due to the nature they are displayed by Notify My Android, or that it was unclear whether they were sent at all. The latter problem was regarded as something that occurred with other applications on the Android system as well. Regarding the content of the trigger opinions varied. One participant claimed that the content of the signal was basically irrelevant, as the notification itself (containing the link to the FWDP) was enough to draw attention to the diary. Each participant preferred a different spark. When asked which trigger they disliked, each participant – again – chose a different one. All participants indicated that the content of the triggers did not provide extra motivation to fill out the diary. One participant did mention in particular that it is important to send the reminders, as filling out the diary is not part of daily routine and can easily be forgotten.
The diary itself performed admirably: no menu choices regarding food type and disposal reason were missed. One participant wondered, however, why the particular options were featured in the FWDP, but could not indicate what options were lacking. One participant did mention that including the “trim waste” option in the reasons menu was a good idea, since it is something that was easily forgotten to add to the diary.
Every participant complained that they missed a way to review previously made diary entries. They wanted to see the result of
their actions or review whether or not they remembered to fill out the diary. One participant said: “Why am I doing this?”
6. DISCUSSION
The data resulting from the experiment clearly shows a positive relation between sending a trigger and filling out the food waste diary, compared to the control group that received no trigger that day. When looking at FWDP visits the first 24 hours after sending the trigger, the social spark performs best, with a normalized ratio of 0.32; followed by the signal and the
hope spark (both 0.21); and lastly: the pleasure spark (0.15).
Not sending a trigger still yields a ratio of 0.11. The pattern is different when looking at the filtered results of 24 hours after sending the trigger: while the social spark (0.28) still performs best, this event result is less pronounced than when looking at visits. The signal (0.23) and hope spark (0.21) from events (24h) perform similarly to visits (24h). The pleasure spark (0.13) now performs worse than no trigger (0.15).
The 3 hour interval yields more pronounced results than looking at the 24 hour interval: the signal performs best, with a 0.34 ratio; followed by the social spark (0.25); and the
pleasure spark and hope spark (both 0.16). Sending no trigger
yields a ratio of 0.09.
From the experiment results it cannot be determined whether the content of the sparks really has an influence on the amount of motivation a participant has on performing the task, compared to using a more neutrally toned signal. It may very well be the case that different messages have a varying effect on different individuals. A fully function system should include user modelling unit, where the system adjusts its behaviour regarding the content of its triggers on the response the triggers generate with its user.
The FWDP contained minimum features only, since the only goal of the experiment was to uncover the influence of triggers on adherence to filling out the diary. It is unclear what the tipping point is to performing the task when more data (such as adding a picture) is requested. One of the generic diary criteria is flexibility, which in this case should mean that, while being able to add more data is nice, it should not be a requirement for a diary entry.
The indication telling the user a diary entry was successfully added was not displayed clearly enough. Participants wondered whether or not the results were processed, they could not retrieve any previous entries and thus wondered if they had already filled out the diary and whether the system worked at all. Future versions should include a way for users to review submitted entries.
Loading times of around a minute to start the prototype is unacceptable. Using a platform (such as Axure) to rapidly produce a prototype can be a boon, but since it outputs clunky code, the environment in which the prototype is used should be controlled.
7. CONCLUSION
This report established the criteria for a system that allows users to keep track of food they dispose of. Existing food waste diaries are examined and compared with criteria flowing from the Fogg Behavioural Model and generic diary design. From these criteria it became apparent that existing systems lacked some kind of trigger mechanism. Whether or not such a mechanism really motivates users to perform a task remains to be seen, but it can be concluded that a trigger is a helpful tool in reminding users to fill out their food waste diary.
8. REFERENCES
[1] E. Commission, “Roadmap to a Resource Efficient Europe.” Brussels, 2011.
[2] E. Commission, “Food Safety - Sustainability of the Food Chain.” [Online]. Available:
http://ec.europa.eu/food/food/sustainability/index_en.ht m. [Accessed: 30-Jan-2014].
[3] Westerhoven, “Bepaling voedselverliezen in
huishoudelijk afval in Nederland, Vervolgmeting 2013,” 2013.
[4] E. Commission, “Second meeting of the working group on food losses/food waste,” no. February. Brussels, pp. 0–17, 2013.
[5] T. E. Quested, a. D. Parry, S. Easteal, and R. Swannell, “Food and drink waste from households in the UK,”
Nutr. Bull., vol. 36, no. 4, pp. 460–467, Dec. 2011.
[6] L. Ventour, The food we waste, vol. 0, no. April. 2008. [7] N. Bolger, A. Davis, and E. Rafaeli, “Diary methods:
capturing life as it is lived.,” Annu. Rev. Psychol., vol. 54, pp. 579–616, Jan. 2003.
[8] B. Fogg, “A behavior model for persuasive design,”
Proc. 4th Int. Conf. Persuas. Technol. - Persuas. ’09, p.
1, 2009.
[9] A. Thieme, S. Lawson, and P. Olivier, “BinCam: Waste Logging for Behavioral Change,” pp. 1–4, 2011. [10] A. Thieme, R. Comber, and J. Miebach, “We’ve bin
watching you: designing for reflection and social persuasion to promote sustainable lifestyles,” Proc. …, pp. 2337–2346, 2012.
[11] “Eurostat - Data Explorer.” [Online]. Available: http://appsso.eurostat.ec.europa.eu/nui/show.do?dataset= ilc_lvph01&lang=en. [Accessed: 12-Jun-2014].
