=Paper=
{{Paper
|id=Vol-2216/healthRecSys18_paper_5
|storemode=property
|title=Nutrilize a Personalized Nutrition Recommender System: an Enable Study
|pdfUrl=https://ceur-ws.org/Vol-2216/healthRecSys18_paper_5.pdf
|volume=Vol-2216
|authors=Nadja Leipold,Mira Madenach,Hanna Schäfer,Martin Lurz,Nada Terzimehic,Georg Groh,Markus Böhm,Kurt Gedrich,Helmut Krcmar
|dblpUrl=https://dblp.org/rec/conf/recsys/LeipoldMSLTG0GK18
}}
==Nutrilize a Personalized Nutrition Recommender System: an Enable Study==
https://ceur-ws.org/Vol-2216/healthRecSys18_paper_5.pdf
Nutrilize a Personalized Nutrition Recommender System:
an enable study
Nadja Leipold∗ Mira Madenach Hanna Schäfer
Technical University of Munich Technical University of Munich Technical University of Munich
Martin Lurz Nađa Terzimehić Georg Groh
Technical University of Munich University of Munich (LMU) Technical University of Munich
Markus Böhm Kurt Gedrich Helmut Krcmar
Technical University of Munich Technical University of Munich Technical University of Munich
ABSTRACT overall occurrence of malnutrition. However, looking at an individ-
A nutrition assistance system gives feedback on one’s dietary be- ual level, people are very different in relation to their dietary needs.
havior and supports behavior change through diverse persuasive This can be due to the phenotypic or genotypic traits of a person,
elements like self-monitoring, personalization, and reflection imple- or the individual diet and lifestyle of that person [5].
mented e.g. with visual cues, recommendations or tracking. While At the same time, mobile applications that support people in
an automated recommender system for nutrition could provide healthier lifestyles reach increasing awareness among society and
great benefits compared to human nutrition advisors, it also faces a industry as well as in research. In combination with intelligent
number of challenges in the area of usability like efficiency, efficacy recommender systems and persuasive designs, they offer a way to
and satisfaction. In this paper, we propose a mobile nutrition assis- face unhealthy lifestyles [20] like unhealthy diets, smoking and
tance system that specifically makes use of personalized persuasive lack of physical activity, that are related to an increasing number of
features based on nutritional intake that could help users to adapt noncommunicable diseases (NCDs) such as cardiovascular diseases,
their behavior towards healthier nutrition. In a pilot study with cancer, chronic respiratory diseases and diabetes [24].
14 participants using the application for 3 weeks we investigate Smartphone applications have already been used as an inter-
how the different features of the overall system are used and per- vention tool (e.g. [3]), but focus mostly on the weight loss of par-
ceived. Based on the measurements, we examine which functions ticipants. There are also several popular commercial weight loss
are important to the users and determine necessary improvements. applications like MyFitnessPal, MyNetDiary and Lifesum. [7] an-
alyzed the most popular mobile applications in this context and
CCS CONCEPTS concludes that they generally lack personalized nutrition with indi-
vidualized feedback as well as nutrition education.
• Applied computing → Health care information systems;
In contrast to these approaches, our nutritional recommender
Health informatics;
system Nutrilize combines personalized recipe recommendations,
KEYWORDS visual feedback and other persuasive measures, as presented by
[21], by considering the personal characteristics and the nutritional
Recommender Systems; Personalization; User Interaction; User Ex- status of 26 macro- and micronutrients.
perience; Nutrition Behavior; enable-Cluster In this paper, we present the characteristics of the Nutrilize sys-
ACM Reference Format: tem as well as a pilot study of this system. We analyze the interac-
Nadja Leipold, Mira Madenach, Hanna Schäfer, Martin Lurz, Nađa Terzime- tion with and perception of this system over a period of 21 days
hić, Georg Groh, Markus Böhm, Kurt Gedrich, and Helmut Krcmar. 2018. considering data from 14 participants.
Nutrilize a Personalized Nutrition Recommender System: an enable study.
In Proceedings of the Third International Workshop on Health Recommender
Systems co-located with Twelfth ACM Conference on Recommender Systems
2 BACKGROUND
(HealthRecSys’18), Vancouver, BC, Canada, October 6, 2018 , 6 pages. This section provides insights into the status of recommendations
in the food domain, in the health domain, in the nutrition science
1 INTRODUCTION domain and within existing applications in general.
In recent years, the need for personalizing dietary recommendations Even though research in the area of food recommendation for
became more and more apparent. Until today, dietary recommen- healthier nutrition becomes more popular due to social relevance,
dations are mostly aimed at the general population to decrease the the number of existing systems is relatively low. [23] as well as [22]
provide state-of-the-art reviews of approaches and systems in the
∗ Email: nadja.leipold@in.tum.de
area of food recommender systems. Various approaches exist to
recommend food and recipes based on different methods that elicit
HealthRecSys’18, October 6, 2018, Vancouver, BC, Canada user preferences using user ratings, scores and tags. For example,
© 2018 Copyright for the individual papers remains with the authors. Copying permit- approaches utilize recipe information and offer recommendations
ted for private and academic purposes. This volume is published and copyrighted by
its editors.
from individual scored ingredients contained within a single recipe
that got formerly rated positively [8] or negatively [12] by users.
HealthRecSys’18, October 6, 2018, Vancouver, BC, Canada N. Leipold et al.
Besides user preferences in certain foods, health becomes more user feedback is primarily based on macronutrients and activity. In-
important as a factor in a food recommendation system due to take tracking or feedback on a micronutrient level, is not considered
the increasing problems with unhealthy eating habits and their re- within the analyzed systems.
lated diseases. Recently, efforts to incorporate health into so-called
health-aware recommender systems have been done by a number of 3 NUTRITION RECOMMENDER SYSTEM
researchers [20]. [10] developed for example a function to derive
the balance between calories needed by the user and contained
by the recipe. [6] addresses the problem of finding the balance be-
tween users’ taste and nutritional aptitude. [23] investigated the
possibility to integrate nutritional facts into their recipe recommen-
dations. Nevertheless, literature on research covering the topic of
incorporating health is limited until now.
There are several national and international dietary guidelines
[17] that provide important standard sources for nutritional infor-
mation. However, they are based on population rather than indi-
Figure 1: Nutrient response curve of the DRI concept [16]
vidual needs. Recent approaches to personalized nutrition show
promising insights into the effectiveness of personalized nutrition
recommendations. For example, [25] investigated individual aspects, To provide meaningful recommendations, we implemented a
which influence the post-prandial glucose response (PPGR) of a knowledge-based, personalized nutrition recommender system.
person to a certain food. They showed, that the PPGR for the same This recommender system relies on four main components: An
meal differs greatly between individuals. Using machine-learning accurate nutritional food database, a user nutrition profile, a recipe
techniques and creating an algorithm based on individual aspects, database, and a knowledge-based utility function for each nutrient.
such as dietary behavior, anthropometrics, blood biomarkers and We compared 3 different sources of food item databases: BLS,
gut microbiome, they were able to accurately predict the PPGR to FDDB and Fatsecret. In the end, we selected the BLS (Bundeslebens-
certain foods. The effectiveness of personalized dietary recommen- mittelschluessel) database [11] due to its high number of accurately
dations for multiple nutrients was also examined in a European represented nutrients. The BLS is used to record the user’s intake
web-based Proof-of-Principle (PoP) study, the Food4Me study [4]. as well as to calculate the recipes nutritional profile. During the
The aim was to compare the effectiveness of personalized nutrition pilot study 26 different micro- and macronutrients were derived
advice (based on dietary, phenotypic and genotypic information) from the BLS for both the user’s intake and the recipes profile.
with population-based advice to improve dietary behavior. In the The user profile has several components. The main influence on
6-months study, personalized dietary advice proved to be more the recommender system is represented by the user’s intake history.
effective than conventional dietary advice in improving nutritional We chose a three-day-average to represent the users nutritional
habits [18]. Food4Me was not solely created for overweight partici- profile. We decided on using an average to avoid contradicting
pants to lose weight, but their main aim was to enhance a healthy advices within one day (e.g. less/more calcium). At the same time,
diet. In [21] we design a mobile system Nutrilize that offers person- we did not want to extend the average further than three days
alized nutrition advice similar to Food4Me and combines it with to be able to react to changes in the users diet. Furthermore, the
new approaches such as recipe recommendations. Nutrilize sup- recommender system integrates gender, age, and BMI to personalize
ports users with recommendations based on the estimated personal the recommendations.
nutritional needs and combines them with principles of persuasion The recipes are obtained from KochWiki1 , which is licensed un-
[19] developed MyBehavior, a mobile application that supports der Creative Commons Attribution - ShareAlike 3.02 . We combined
users with different personalized feedback in terms of actionable the recipe database with the nutritional information for each food
suggestions. These are based on algorithms from decision theory item in the BLS database using an adaptation of [13]. Overall, 240
that learn users’ physical activity and dietary behavior. They in- recipes are provided during the study.
clude users’ preferences as well as behavioral change strategies to For the recommendations, each recipe is rated by comparing its
give appropriate personalized feedback on diet and physical activity. nutritional profile with the nutritional needs of the user. The user’s
Besides scientific approaches, commercial food diaries and/or diet needs are derived using the dietary reference intakes (DRI) from the
coaches with incorporated physical activity trackers, mainly focus- Institute of Medicine [15] and from the D-A-CH reference values [9].
ing on reduction of calorie intake such as MyFitnessPal, MyNetDi- The dietary reference intake [16] is divided by age and gender and
ary, Lifesum, etc. offer various forms of visual and textual feedback structured as shown in figure 1. For the purpose of estimating the
(e.g. overview charts on calorie intake and expenditure, and the nutrient intake status of a person, intakes below the EAR (estimated
macronutrients’ distribution of consumed foods). According to a average requirement) are categorized as insufficient intake, intakes
review on nutrition-related mobile applications in the UK [7], the above the UL (upper limit) are categorized as a likely overdose, and
analyzed applications lack personalization and educative aspects. intakes between EAR and RDA (recommended daily allowance) are
Partially, they include individual aspects like age, gender, weight categorized as possibly insufficient intake, while intakes between
and other phenotypes. However, the information used to generate the RDA and UL are categorized as optimal intake. Based on these
1 www.kochwiki.org
2 https://creativecommons.org/licenses/by-sa/3.0/
Nutrilize a Personalized Nutrition Recommender System HealthRecSys’18, October 6, 2018, Vancouver, BC, Canada
reference functions, the user’s needs are described as a vector of of the current nutrient status. Feedback calculations here are based
26 advice values. To derive a recipes utility (u) to improve a user’s
nutritional profile, the nutrient profile of the recipe (r) is multiplied
with the need/advice profile of the user (a), resulting in a rating
score. During this multiplication, some nutrients (pi ) are weighted
(w) higher based on certain input parameters of the participant:
rp1 wp1 ap1 ur,p1
. . . .
.. ◦ .. ◦ .. = .. (1)
r w a u
pn pn pn r,pn
Finally, all recipes are ranked per meal by the sum of their ratings
and shown to the user. In addition to the recipes, the users received
an explanation on which nutrient influences the ranking of this
recipe the most and which benefits this nutrient provides.
Figure 3: Nutrient details screen (l), nutrient overview (m)
4 NUTRILIZE INTERFACE DESIGN
and statistics overview (r)
The developed mobile smartphone application, which is used for
this study, is based on the intervention tool presented by [21]. It on the average of the three previous days of consumption. The
consists of three main components in terms of a food diary, visual six most critical nutrients (regarding the highest aberration from
feedback and recipe recommendations. the suggested intake amount) are shown. The color coding used
in the application consists of a traffic light color scheme that pro-
4.1 Food Diary vides a high association for the users [2]: red (for warnings), yellow
(for attention) and green (for go on). In case of optimal behavior,
even the six most critical nutrients would show a green symbol.
Additionally, the arrows in the circles in the home screen indi-
cate recommended behavior (pointing up: increase intake; pointing
down: reduce intake). On the bottom of the home screen we added
four circular buttons for easy diary access to add new meals. When
using the sports button, the user can fill out a questionnaire to esti-
mate the physical activity level [14]. Finally, users can access their
recommendations through the white button on the home screen.
Through clicking on a nutrient on the home screen, an informa-
tion page is shown (Figure 3, left). There, the current nutrient status
is visualized via a colored horizontal bar, showing the current value
as a blue vertical line and the areas of intake represented with the
same color coding as in the home screen. Furthermore, the intake
Figure 2: Diary (l), home screen (m) and food search (r) development over the last three days is visualized. In addition to
the visual feedback some information is given in textual form, such
In order to provide personalized feedback and recommendations, as information on the nutrient, its importance for the human body
the application needs regular input of the user’s nutrition behavior. and possible adverse effects caused by over- or under-consumption.
This can be tracked via the integrated personal food diary supplied Below the nutrient description, the main food sources for this nu-
by nutritional information from the BLS database (Figure 2, left). trient are listed as well as the personalized reference values for the
We added the meal categories "Breakfast", "Lunch", "Dinner" and consumption of this nutrient.
"Snacks" for better structuring. The diary can be filled by clicking By clicking on the middle circle in the home screen, the user
the plus button at each diary section or by using the shortcut on can access the personal nutrition overview (Figure 3, middle). It
the home screen (Figure 2, middle). When adding food to the diary, lists all 26 nutrients with their current status, visualized through a
a search dialog is opened (Figure 2, right), where users can search horizontal bar as on the nutrient detail screen. Users can further-
their meals in the database. After selecting a result, the user can more access detailed statistics on their previous nutrition behavior
adjust the amount of the food item before adding it into the diary through the applications menu (Figure 3, right). This visualization
or change the amount afterwards in the diary view. For the purpose allows the user to see the progress within a week or a month.
of a quick access of previous chosen meals and related quantitative
disclosures the user is offered a Recent tab below the search bar. 4.3 Recipe Recommendations
The recipe recommendations offer ranked lists of recipes (as de-
4.2 Visual Feedback scribed in section 3) for each meal, based on their nutrient content
Information graphics are generated for different visual feedback and the user’s nutritional history of the last three days. They are
screens. The home screen (Figure 2, middle) provides an overview provided in separate tabs for each of the four meal categories, as
HealthRecSys’18, October 6, 2018, Vancouver, BC, Canada N. Leipold et al.
shown in Figure 4. The traffic light color scheme is used here as well participation. Out of 31 participants, who finished the first screen-
and represents the overall "health benefit" of the recipe according to ing, 20 were both suitable for participation and finished the first
the user’s current nutrition status. Each recommendation consists survey. The final survey was concluded by 18 participants. Overall,
of a recipe title, a picture and a coarse overview of the recommended only 14 of the 20 participants concluded all measurements. Those
amount and relative content of macronutrients. Additionally, users 14 users are further examined in this paper.
can click on the explanation button to receive insights into why
this recipe is recommended to them. 5.2 Measures
All additional information on the recipe, such as a detailed list We had three different types of measurements in this study. First,
of ingredients and the preparation instruction can be viewed when we measured the nutrient intake of participants. In the beginning
clicking on the recipe item within the list (figure 4). The users can and end we derived the users’ dietary intake from a food frequency
view the ingredient list for one portion or with the recommended questionnaire using 150 common food items. Afterwards, we let
sizes for the user (based on their caloric requirements). They can the participants track their nutrition within our application for 21
immediately add the consumed portion of a recommended recipe days. Based on their input, we were able to derive daily nutritional
to their diary, saving the time of entering each single ingredient. information. Second, we measured the participants’ usage behav-
ior within the application using an open analytics and tracking
tool named Matomo 4 , formerly Piwik. The tracking tool allowed
us to measure the time and number of actions within each appli-
cation session. It furthermore tracked predefined goals, such as
accepting a recommendation. Third, we measured the participants’
self-reported attitudes and perceptions. In a pre-study questionnaire
we asked them about their background, cooking habits, their health
attitude, and their technology attitude. In a post-study survey, we
assessed the overall usability using a System Usability Scale (SUS)
questionnaire [1] and specific feedback for each application feature.
6 STUDY RESULTS
This section shows the results of our user study for the different
measurements. First, we look at the characteristics of the study
Figure 4: Recommendation list (l) and recipe screen (r) group. Then we analyze the system perception by the participants
and how they used it during the study. Finally, we analyze the
nutritional data retrieved from both the application’s diary and the
5 USER STUDY food frequency questionnaires. Our goal is to get an understanding
This study represents an exploratory pilot study of the Nutrilize of the needs of our participants, the effects of the system and the
system. We focused on study group, system interaction, system required changes for the system.
perception and reported dietary behavior. The study protocol was
approved by the ethical committee of the Faculty of Medicine of 6.1 Study Group
the Technical University of Munich in Germany (no. 477/16 S).
Table 1: User characteristics of 14 participants. Health and
5.1 Study Procedure technology attitude are measured with 6 questions each on
Participants were recruited from the enable research participation a 5 point Likert scale (0 disagree - 5 agree)
database 3 with approx. 120 invitations. The study consisted of
four distinct steps. First, all participants completed a screening Age Height Weight BMI Health Tech. Tech.
questionnaire that checked for medical (e.g. allergies, pregnancy, Attitude >=50y <50y
etc.) and technical constraints (e.g. Android phone, Internet access, Min 23 152 52 18,4 3,3 1,8 2,8
etc.). Second, if participants matched study constrains and gave Max 65 183 113 36,1 4,5 3,5 5,0
their consent, they received a link to the first survey (time point Avg 45 170 77 26,6 3,9 3,0 4,3
0). In this survey, we collected data on dietary habits using a food
frequency questionnaire (FFQ), on activity habits using the Nor-
man questionnaire [14] and on their anthropometric measures. The Table 1 shows the user characteristics, the health attitude, and
anthropometric measures included self-measurements of the body the technology attitude of the participants below and above an
height, bodyweight and waist/hip circumference. Third, one day age of 50 years. The gender ratio is slightly biased with 8 female
after the first survey all participants received the Nutrilize applica- and 6 male participants. This tendency is lower than expected.
tion and an instruction manual. Fourth, after 3 weeks of using the The balance can be explained by the recruitment target, which is
application, the participants received the final survey (time point 3) already balanced and interested in healthy nutrition in general.
asking for feedback on the system. They received no payment for The age of the participants ranges from 23 to 65 years. With an
3 http://enable-cluster.de/index.php?id=198&L=1 4 https://matomo.org/
Nutrilize a Personalized Nutrition Recommender System HealthRecSys’18, October 6, 2018, Vancouver, BC, Canada
average age of 45 years, the study group is significantly older than analyze the caloric behavior during the study and third, we consider
expected. In part this leads to different attitudes towards technology the health measurement represented by a sufficient nutrient intake
in general. We furthermore see a full range of BMIs. One participant according to the reference guidelines. The eating and input tracking
is underweight, five participants are of normal weight, four are habits can be viewed by looking at the different consumed meals.
overweight, and three are obese. Finally, all participants have very The data shows that breakfast and dinner are very similar with an
similar health attitudes. The majority (12 out of 14) feels unwell with average number of 70 tracked food items during the 21 days. Lunch
their current diet, but they believe they can keep up the changes is only reaching 58 tracked items. This difference might be due to
required of them, even if a transition would be difficult. the limited time for systematic tracking during lunch. Finally, the
snack category was used very differently amongst the participants
6.2 System Interaction and Perception ranging from 0 to 66 items.
The first aspect of the system perception is the overall usability
of the application. The SUS [1] feedback resulted in a score of 52,
which shows that the application is not a basic prototype anymore,
but also not on an average usability level yet. Next, we tracked the
user’s interactions within the application. Most of the interactions
(on average 85%) are focused on filling in the dietary diary of the
users. Among the features the application is offering to the users,
the patterns are less uniform. As visible in figure 5 some users
prefer the visual feedback given in retrospective (user 1,5,10), while
others focus almost entirely on the recommendations (user 2). A
few users are even putting their emphasis on the calorie overview
(user 12). During the final questionnaire, the feedback shows that
Figure 6: Comparison of reported daily energy intake (kCal)
calculated based on FFQ (green) or on the application based
dietary tracking (blue)
The caloric behavior during the study shows that the intake
tracked within the application is systematically lower than the one
derived from the FFQs. In figure 6 the tracked intake is shown as a
box plot for every day (blue). Additionally, the measurement with
the standardized food frequency questionnaire before and after the
study is shown. For most participants, the daily intake is about 1000
kCal lower when tracking with the application. At the same time,
the calories calculated from the FFQs stay in a similar range with a
Figure 5: Percentage of interactions within the application slight tendency to less intake after the study.
The healthiness of nutrition in this study is defined by the num-
the system still needs to improve. The single nutrient visualizations ber of adequately consumed nutrients over the past three days,
are perceived very well for all visual representations. Other features which were calculated and presented as nutrient intake per day.
however, such as the recommendations, were not perceived as well. The highest number of optimally ingested nutrients (22 out of 26
In general, feedback on the recommended recipes included missing nutrients) is reached by one participant after 11 days of interven-
variety, difficulty of recipes, and missing personal adaption of the tion. The average number of adequately ingested nutrients is 13,
recommendations (e.g. raw food or vegetarian). The explanations which is only half of the tracked nutrients. This might in part be
within the recipes on the other hand were perceived as helpful caused by the underestimation of food intake.
by most participants, possibly because they link back to focusing
on the single nutrients from the visualizations. Finally, the diary 7 DISCUSSION OF RESULTS
function of the application was clearly (12/14) preferred over the One of the main challenges, that can be drawn from the results is
FFQ input method, even though it is more time consuming. usability. As working with a prototype system this is not surprising.
Besides the current perception, we also inquired the wishes for Nevertheless, important next steps could be extracted from the
our future system. The users suggested easier entry methods for feedback, which are crucial for an improved usability and for an
the food diary, detailed sports tracking, greater recipe variety, more application which is supposed to be used daily.
positive feedback and better general performance and design. The first feature that should be improved are the recommen-
dations. Although we included a recommendation system, which
6.3 Nutrition Behavior produces highly personalized and individual recommendations,
The nutrition behavior can be analyzed on different levels. First, the users are facing many constraints in real life situations that
we can look at the eating habits of the participants. Second, we can were not modeled. These factors include the availability of certain
HealthRecSys’18, October 6, 2018, Vancouver, BC, Canada N. Leipold et al.
food (e.g. seasonal fruits), personal preferences (e.g. vegan) and REFERENCES
group constellations (e.g. a mother who should cook for her family). [1] J. Brooke. 1986. System usability scale (SUS): a quick-and-dirty method of system
Moreover, additional recipes seem to be necessary since the current evaluation user information. Reading, UK: Digital Equipment Co Ltd (1986).
[2] C.M. Brown. 1998. Human-computer interface design guidelines. Intellect Books.
recommendations were often perceived as repetitive. [3] M.C. Carter, V.J. Burley, C. Nykjaer, and J.E. Cade. 2013. Adherence to a smart-
An additional learning for the personalized recommender system phone application for weight loss compared to website and paper diary: pilot
randomized controlled trial. Journal of medical Internet research (2013).
is the high dependency of some advice functions on accurate user [4] C. Celis-Morales, K.M. Livingstone, and C.F.M. Marsaux et al. 2015. Design
input. When some meals are tracked or the amount of a food item and baseline characteristics of the Food4Me study: a web-based randomised
is underestimated (which seemed to be a trend in the pilot study), controlled trial of personalised nutrition in seven European countries. Genes &
nutrition (2015).
the users do not reach the recommended consumption values of [5] C. Celis-Morales, K.M. Livingstone, and C.F.M. Marsaux et al. 2016. Effect of
the macronutrients. This can result, for example, in suggestions to personalized nutrition on health-related behaviour change: evidence from the
increase the user’s intake of fat and thus providing recipe sugges- Food4me European randomized controlled trial. International journal of epidemi-
ology 46, 2 (2016), 578–588.
tions of high fat foods. To prevent such inverse advice, we suggest [6] D. Elsweiler, M. Harvey, B. Ludwig, and A. Said. 2015. Bringing the" healthy"
excluding total amounts of carbohydrates and fat while including into Food Recommenders. In DMRS.
[7] R.Z. Franco, R. Fallaize, J.A. Lovegrove, and F. Hwang. 2016. Popular Nutrition-
proportional advice on specific types of fat and on sugar. Related Mobile Apps: A Feature Assessment. JMIR mHealth and uHealth (2016).
Besides improvements, the users also reported the missing of [8] J. Freyne and S. Berkovsky. 2010. Intelligent food planning: personalized recipe
some functionalities. For example, some users wished for the ability recommendation. In Proceedings of the 15th international conference on Intelligent
user interfaces - IUI ’10.
to track their physical activity manually (instead of with a question- [9] D-A-CH (Deutsche Gesellschaft für Ernährung Österreichische Gesellschaft für
naire). This would suggest a sports diary comparable to the current Ernährung Schweizerische Gesellschaft für Ernährungsforschung Schweizerische
food diary with the option to integrate data from popular fitness Vereinigung für Ernährung). 2008. Referenzwerte für die Nährstoffzufuhr. Umschau
Braus Verlag.
trackers. Furthermore, the home screen was perceived to be dis- [10] M. Ge, F. Ricci, and D. Massimo. 2015. Health-aware Food Recommender System.
couraging. Some UI changes might easily improve this perception. In Proceedings of the 9th ACM Conference on Recommender Systems.
[11] B.M. Hartmann, S. Bell, A.L. Vásquez-Caicedo, A. Götz, J. Erhardt, and C. Brom-
One possibility could be to show a progress on the optimization of bach. 2005. Der Bundeslebensmittelschlüssel. German Nutrient DataBase. Karl-
nutrients in the center circle on the home screen. sruhe: Federal Research Centre for Nutrition and Food (BfEL) (2005).
Finally, the high amount of time spent on the intake tracking [12] M. Harvey, B. Ludwig, and D. Elsweiler. 2013. You are what you eat: Learning
user tastes for rating prediction. (2013).
(85%) might offer a chance. Instead of giving only support for ret- [13] M. Müller, M. Harvey, D. Elsweiler, and S. Mika. 2012. Ingredient matching to
rospective or perspective actions, some visual cues might be inte- determine the nutritional properties of internet-sourced recipes. In Pervasive
grated within the action of tracking itself. Furthermore, the sub- Computing Technologies for Healthcare (PervasiveHealth), 2012 6th International
Conference on.
sequent implemented system addresses this issue by offering fast [14] A. Norman, R. Bellocco, A. Bergström, and A. Wolk. 2001. Validity and repro-
access to favorite foods for use in the longitudinal study. ducibility of self-reported total physical activity-differences by relative weight.
International journal of obesity (2001).
Overall, the system still needs some improvement, but already [15] Institute of Medicine (US) Subcommittee on Interpretation and Uses of Dietary
43% of users stated that they would use the system frequently. Of Reference Intakes; Institute of Medicine (US) Standing Committee on the Scien-
the 14 participants 9 used the system for more than 17 days. This tific Evaluation of Dietary Reference Intakes. 2000. DRI Dietary Reference Intakes:
Applications in Dietary Assessment. Washington (DC): National Academies Press
shows that the general idea and purpose of the system are relevant (US).
to the study group. However, further adjustments might increase [16] J.J. Otten, J.P. Hellwig, and L.D. Meyers et al. 2006. Dietary reference intakes: the
the number of users willing to use the system frequently and might essential guide to nutrient requirements. National Academies Press.
[17] J. Painter, J.-H. Rah, and L. Yeon-Kyung. 2002. Comparison of international food
consequently create an effective tool for nutrition behavior change. guide pictorial representations. Journal of the Academy of Nutrition and Dietetics
(2002).
[18] R. Poinhos and M.D.V. de Almeida. 2015. Personalised nutrition: paving a way
8 CONCLUSION AND FUTURE WORK to better population health (A White Paper from the Food4Me project). Technical
Mainly, this study shows the need for improvements in several as- Report.
[19] M. Rabbi, M.H. Aung, M. Zhang, and T. Choudhury. 2015. MyBehavior: auto-
pects of the application, such as the recommendations, performance matic personalized health feedback from user behaviors and preferences using
and ease of intake tracking. Still, 43% of the participants would use smartphones. In UbiComp.
the application regularly and most (85%) prefer the daily dietary [20] H. Schäfer, S. Hors-Fraile, R.P. Karumur, A. Calero Valdez, A. Said, H. Torkamaan,
T. Ulmer, and C. Trattner. 2017. Towards Health (Aware) Recommender Systems.
tracking to a weekly food frequency questionnaire. Some major In Proceedings of the 2017 International Conference on Digital Health - DH ’17.
challenges remain open, such as integrating contextual and social [21] N. Terzimehić, N. Leipold, H. Schäfer, M. Madenach, M. Böhm, G. Groh, K. Gedrich,
information as well as the accuracy of the received input data. In and H. Krcmar. 2016. Can an Automated Personalized Nutrition Assistance
System Successfully Change Nutrition Behavior? - Study Design. In Thirty Seventh
future, we will improve Nutrilize according to the given feedback International Conference on Information Systems.
and evaluate the long term (3 months) effect of using the applica- [22] T.N. Trang Tran, M. Atas, A. Felfernig, and M. Stettinger. 2018. An overview
of recommender systems in the healthy food domain. Journal of Intelligent
tion against a control group. Finally, this works provides starting Information Systems (2018).
points about the integration of nutritional recommender systems [23] D. Trattner and D. Elsweiler. 2019. Food Recommender Systems: Important Con-
into more holistic persuasive mobile systems for daily usage. tributions, Challenges and Future Research Directions. In Collaborative Recom-
mendations: Algorithms, Practical Challenges and Applications, Shlomo Berkovsky,
Iván Cantador, and Domonkos Tikk (Eds.). World Scientific Publishing.
ACKNOWLEDGMENTS [24] WHO. 2014. Global status report on noncommunicable diseases 2014. World
Health (2014).
The preparation of this paper was supported by the enable Clus- [25] D. Zeevi, T. Korem, and N. Zmora et al. 2015. Personalized Nutrition by Prediction
ter and is cataloged by the enable Steering Committee as enable of Glycemic Responses. Cell (2015).
025 (http://enable-cluster.de). This work was funded by a grant
of the German Ministry for Education and Research (BMBF) FK
01EA1409A.