Martin's Blog – Research on Human-Computer Interaction and Ubiquitous Computing

Featured

Productive, Anxious, Lonely – 24 Hours Without Push Notifications

Imagine, all notifications would go silent. No more buzzing, flashing, beeping, or pop-ups.

This is exactly the situation that we created in the Do Not Disturb Study. The study was conducted as part of a collaboration between the Scientific Group of Telefonica R&D and Human-Computer Interaction Institute at Carnegie Mellon University, Luz Rello and I envisioned the Do Not Disturb Challenge.

The Do Not Disturb Challenge

We asked 30 volunteers to disable notification alerts for 24 hours across all devices and all services. We carefully walked the participants through all devices and services. Where possible, we used system-wide settings, such as the Do Not Disturb mode, to suppress all alerts. In other cases, such as Skype, we showed people how to disable notifications in the settings of the respective app. Please note that participants could still read messages or emails, they would simply not receive any alert.

To study the effect that notifications have on us, we captured self-reported feedback, and compared it to the same self-reported feedback collected via questionnaire during a normal baseline day. Furthermore, after the study, we conducted a post-hoc interview to uncover themes that we had not anticipated. We discovered the following main effects.

Drop in Engagement and Reduced Responsiveness

The absence of notifications had a significant effect on how participants perceived their engagement with the mobile phone. For example, Participant #02 “forgot my phone at work” because of not being reminded of the phone by notifications.

Increased Productivity

As expected, notifications distract. Hence, the answers of the questionnaire show that participants felt significantly less distracted and more productive: Participant #07 said that it was “easier to concentrate, especially when working on the desktop (computer).”

Lack of notifications caused to miss information

During the day without notifications, participants were significantly more likely to agree with the statements that they missed professional or personal information. During the post-hoc interview, we collected several anecdotes. For example: because of the lack of notifications, Participants #12 forgot to continue a chat with a friend. As a consequence, this friend got angry for not receiving replies.

Lack of notifications induced worry

Consequently, participants were significantly more likely to agree to the statement “I felt worried about missing notifications“. For example, Participants #04 “was meeting with [a friend] for lunch, and I knew that I was going to receive something from her“.

More frequent checking of the phone

During the day without notifications, agreement to the statement “I frequently turned on the phone to check for missed notifications“. For example, Participant #12 stated that “because of the reaction of my friend, who got angry because I forgot to respond, I was the whole afternoon with phone in my hand.”

Stress

Interestingly, there were no significant effects on the two stress related items, neither on “I felt stressed” nor on “I felt relaxed“. This might be explained by the finding that there are two opposing stress-inducing effects at work — stress from the interruptions and stress from being anxious to miss important information or violate expectations –, which influenced participants to different extents.

Reduced feeling of social connectedness

Our study revealed a link between notifications and staying emotionally in touch with one’s social group. During the day without notifications, agreement to the statement: “I felt connected with my social group” was significantly lower. These results contrast that — while work-wise, disabling notifications helped to be more focused and productive — socially, they negatively affect the feeling of being in touch with one’s social group.

Polarized reactions to being without notifications

The participants’ post-study reflections to having notifications disabled varied greatly. They ranged from very positive responses, such as “It was amazing! I felt liberated! (Participant #22) over neutral responses, such as “It was not a big deal, since I am usually not checking notifications and people know that I am not responsive” (Participant #25) to very negative responses, such as “I was paranoid and I even left the screen on not to miss a friends notification“} (Participant #04).

The main predictor for the participants’ attitude that we observed was to what extent others typically expected them to respond quickly to messages: the faster the usual response, the more negative the experience.

Signs of notifications overload

For more than two-third of the participants, the participation in the Do Not Disturb Study caused them to reflect on their notifications usage. Almost half of the participants stated the plan to use Do Not Disturb or similar similar notification-suppression modes in the future. For example, Participants #24 realized that “when I need to really get things done, I need to turn notifications off.”

One third stated the plan to manage notifications more consciously. For example, Participants #20 was “considering to only keep notifications for the important things, so people can better reach me“. Participants #26 had come to the conclusion that the “important apps are Messenger, Hangout and WhatsApp.” This shows how important instant messaging has become: people depend on notifications to maintain the expected level of responsiveness. This also shows that – despite the negative effects of notifications – disabling them altogether is not an option.

Two years later, we contacted the 22 participants who intended to manage notifications differently in the future. More than 75% of the participants had followed or followed partially through with their plans.

The fact that more than half of the participants reduced the number of notifications that they are exposed to on a daily basis is a warning sign that our participants were realizing a sense notification overload.

Conclusions

In conclusion, our results show strong and polarized reactions to being without notifications:

Notifications negatively impacted focused work, as participants reported to feel significantly less distracted and more productive without them. This is evidence that disabling notifications can have positive effects.

At the same time, disabling notifications also had significant negative effects: it made participants more worried to miss important information, not being responsive enough, and feeling less connected with their social network. Thus, disabling notifications altogether is not an option.

In contrast to a previous deprivation study, where all participants re-enabled work email notifications after the study, about one-third of our participants expressed the intention to disable some sources of notifications, and about half of our participants expressed the intention to use Do Not Disturb (and equivalent settings) more often in the future. Two years later, 60% of these participants are still following through with their intentions. Another 18% have changed their notification-related behavior.

TL/DR: Notifications. Can’t live with them, can’t live without them.

The work will be presented at ACM MobileHCI ’17, the 19th International Conference on Human-Computer Interaction with Mobile Devices and Services, which will take place in September 2017 in Vienna, Austria.

Citation:
Productive, Anxious, Lonely – 24 Hours Without Push Notifications.
Martin Pielot and Luz Rello.
MobileHCI ’17: ACM International Conference on Human-Computer Interaction with Mobile Devices and Services, 2017.

Share this:

Too Tense for Candy Crush – How Emotions Affect what Kind of Distractions we Welcome

Push notifications are increasingly being used to engage mobile device users with app content. News organizations deliver breaking-news notifications, social platforms inform about new content, games inform about status updates game, etc … with the goal of making the user engage with the service.

In this research, we – Kostadin Kushlev, University of Virginia, Bruno Cardoso, KU Leuven, and myself – explored to what extent users’ current affect, that is, how they are feeling, impacts user engagement. To this end, we analyzed data from a study conducted by Telefónica Research where the participants (N = 337) downloaded a custom-developed app that delivered notifications.

After attending to a notification (N = 32,704), participants reported how they felt in a mini questionnaire. Besides asking how the participants felt, the questionnaire also offered them to voluntarily engage with further content. Participants were not aware that we our main interest was in observing their interaction with said content — they believed that it was mainly there as a courtesy to make their participation in the study more fun.

Participants always had two choices: a mentally demanding and a simple/diverting task. The tasks in these groups were chosen from a list of 4 options each. The mentally demanding offers included: browsing trending games on Google Play, reading the Wikipedia article of the day, filling out a personality questionnaire, or playing a thinking game. The simple and diverting option included watching a trending video, reading fun facts, playing an action game, and watching trending gif images.

The results show a clear impact of affect on the choice of the content:

When feeling good, people tend to avoid mentally demanding tasks. Hence, proactive recommendations for content that requires mental effort should target moments of neutral or even negative valence.
When tense, people tend to avoid diverting tasks. Thus, people who want to reduce task-induced stress might want to rely on external timers to schedule regular breaks with fun activities.
When energetic, people tend to avoid suggestions for further distraction altogether. Hence, proactive recommendations should target moments of low energetic arousal, such as moments of boredom.

These findings show that the current emotional state affects the kind of content users choose to engage with. Future “smart” devices should not only be technologically smart, but also psychologically smart. They should strive to understand how users feel in order to engage them with the most appropriate content at the most opportune of times.

Citation:

Too Tense for Candy Crush: Affect Influences User Engagement With Proactively Suggested Content.
Kostadin Kushlev, Bruno Cardoso, Martin Pielot.
MobileHCI ’17: ACM International Conference on Human-Computer Interaction with Mobile Devices and Services, 2017.

Share this:

Acceptance of Highly-Personalized Ads

Users often perceive ads annoying (when they are unrelated to their interests), or on the other hand they find ads creepy or scary (when matched to their interests and activities). Assuming the recommendation systems will become able to better and better match ads to our interests, how will the users react?

Many apps and internet services are free-to-use. The cost of developing and running these services is paid by bringing advertisement to our attention. However, ads are often annoying an unrelated to our interests.

Hence, services collect more and more information about us to better target and personalize those ads. While research shows that personalized ads can be less annoying and more effective, they require the logging of our demographics and our behavior, such as our searches, the content we browse, or our whereabouts and movements.

Personalization has raised concerns: users worry about the personal data that is being used to create personalized ads. In previous studies, when given a choice, many people expressed hesitation to share information: their concerns of sharing personal data –such as browsing and location history– outweigh the perceived usefulness of personalized ads.

If such concerns persist, investing effort into further personalizing ads would not be worthwhile, as people would not accept them.

We therefore set out to explore the research question: “would people be willing to share their personal data in exchange for highly-personalized online ads?”

To answer this question, we conducted a so called Wizard-of-Oz deception study, i.e., a study in which we simulated a system that can generate highly-personalized ads. Our volunteers were exposed via a web browser to three different highly-personalized ads, designed by people who knew them well. They were made believe that the ads had been generated automatically by an Artificial Intelligence engine on the basis of their browsing & location history and/or personal traits.

The participants’ reactions were surprisingly favorable:

in more than 50% of the cases, the ads triggered spontaneous positive emotional reactions;
almost 90% of participants would share at least two of the three data sources with advertisers; and
about 50% would share all data sources.

Our results provide evidence that highly-personalized ads may offset the concerns that people have about sharing their personal data. Thus further efforts in building increasingly personalized online ads would represent a worthwhile endeavor.

The work was presented at ACM UMAP ’17, the ACM International Conference on User Modelling, Adaptation and Personalization, which took place in July 2017 in Bratislava, Slovakia.

Citation :
“OMG! How did it know that?” Reactions to Highly-Personalized Ads.
Aleksandar Matic, Martin Pielot, Nuria Oliver
UMAP ’17 (Adjunct): ACM International Conference on User Modelling, Adaptation and Personalization, 2017.

(pdf)

Share this:

Optimal Font Size for Web Pages (CHI 2016)

What is the optimal font size for the body of websites?

Guidelines often suggest 10, 12, or 14 points. 12 points is a number you see frequently mentioned in forums and blogs.

In 2009, Smashing Magazin found that a set of 50 very popular news pages used 13px on average. In 2013, they repeated the study and found that 14px and 16px had become the most popular font sizes. In 2015, Firefox and Chrome ship with a default of 16 pixels. However, guidelines and recommendations rarely cite scientific evidence.

Beware, though, points and pixels are not the same! Points are resolution independent. 72 points equal 1 inch. 16 pixels ususally equal 12 points.

Scientific research has been comparing font sizes 10, 12, 14 points and repeatedly found that bigger font implies better readability. This indicates that the current development of increasing font sizes in moving in the right direction. However, the larger the font becomes, the less text fits in one line. Is there a point where fonts become too large?

We were not able to find any scientific study that studies font sizes beyond 14 points. Thus, it was not clear to what extent increasing font sizes beyond 14 points improves readability.

We present results from the first, published scientific including font sizes 18, 22, and 26. In brief, the evidence suggest that readability keeps improving for larger fonts: 18 and 22 points.

Make it Big! The Effect of Font Size and Line Spacing on Online Readability. from Martin Pielot

We, that is with first author Luz Rello, who is currently affiliated with Carngie Mellon’s Human Computer Interaction Institute, co-author Mari-Carmen Marcos, from the Universitat Pompeu Fabra, and me, just published results from a study at ACM CHI 2016, the premier scientific conference on human factors.

The study provides evidence that objective readability (measured via eye-tracking) and comprehension of texts (number of correctly answered comprehension question) significantly improves when using larger-than-usual font sizes (18, 22, 26 points).

In this study, 104 people read Wikipedia articles with different font sizes (10, 12, 14, 18, 22, 26 points – within-subject factor) and line spacings (0.8, 1.0, 1.4, 1.8 – between-group factor) while their reading was recorded with an eye-tracker.

From the eye-tracking data, we extracted the mean fixation duration of the eye movements, which is an established proxy for objective readability. When reading, the eye does not move continuously over the text. It alters between short fixations and saccades. The shorter those fixations are, the less difficulties the reader encounters, which means that the easier to text is to read.

Mean fixation duration per font size. Lower fixation durations mean that the text is easier to read.

In our data, the mean fixation duration dropped absolutely continuously with increasing font size until 18 points. The shortest mean fixatution durations were recorded for 22 points. Subjective readability was best for 18 point font size.

Further, for each text, the participants had to answer comprehension questions. The fraction of correctly answered comprehension questions was significantly lower for font sizes 10 and 12 points. This is impressive evidence about how small font sizes impair readability to an extent where comprehension gets affected.

Fraction of correctly answered comprehension questions per font size condition

In the study, we also tested different line spacings (0.8, 1.0, 1.4, and 1,8). However, line spacing had only minimal effects: we found weak evidence that extreme line spacings (0.8 and 1.8) may impair readability and comprehension.

In summary, our work supports recent calls for drastically increasing font size of website bodies. Our recommendation is to use 18 points font size and default (1.0) line spacing. Just to be clear, in a standard desktop setting, you need to set your Firefox or Chrome browser to 24 pixels to achieve this. In our data, this configuration strikes the balance between having the best readability, comprehension, subjective perception scores, and allowing to fit as much text on the screen as possible.

Of course, this recommendation does not consider aethetic aspects. It is simply about maximizing readability and comprehension of websites.

The work will be presented at ACM CHI ’16: the premier Conference on Human Factors in Computing Systems in May 2016 in San Jose, CA.

Citation:

Make It Big! The Effect of Font Size and Line Spacing on Online Readability.
Luz Rello, Martin Pielot, and Mari-Carmen Marcos.
CHI ’16: ACM SIGCHI Conference on Human Factors in Computing Systems, 2016.

Share this:

Are You Prone to be Bored? Your Phone Can Tell

This work is a follow-up project on our research on When Attention is not Scarce – Detecting Boredom from Mobile Phone Usage

People using their mobile phone in the metro to kill time.

We might think that technology has solved the problem of boredom. More and more devices provide us with an ample source of entertainment at our fingertips.

Paradoxically, today we appear to be more prone to boredom than ever before. The explanation might be that over time people habituate to an increasing exposure to stimuli such that, when the level of stimulation drops, they become bored.

In an extension on our study of detecting phases of boredom from mobile phone usage, in this study, we (Aleksandar Matic, Nuria Oliver, and me of the Scientific Group of Telefonica) explored to what extent technology use is intertwined with boredom proneness, and whether the level of boredom proneness can be inferred from it. We collected data on the accumulated daily mobile phone usage patterns of 22 volunteers, such as, the average number of apps started in a day or the variance of the amount of notifications received per day. Then, those participants filled our the standardized Boredom Proneness Scale.

We found that daily usage patterns can estimate whether the person is above-average prone to boredom with an accuracy of over 80%. Individuals with high boredom proneness were having more unstable daily phone usage patterns: they launched a higher number of apps per day, had strong peaks of social network activity, and turned on the phone a lot. However, surprisingly, the overall time of using the phone was not higher than for individuals with lower boredom proneness.

Boredom proneness is related to a number of negative outcomes, such as depression, drug & alcohol consumption, or anxiety. Obtaining boredom proneness in an unobtrusive, automatic way can, amongst other things, help in the adjustment of the treatments of such health issues.

The work was presented in September 2015 at the ACM International Joint Conference on Pervasive and Ubiquitous Computing, which took place in Osaka, Japan.

The details of the work are described in
Boredom-Computer Interaction: Boredom Proneness and The Use of Smartphone
Aleksandar Matic, Martin Pielot, Nuria Oliver
UbiComp’ 15: ACM International Joint Conference on Pervasive and Ubiquitous Computing, 2015.

Share this:

When Attention is not Scarce – Detecting Boredom from Mobile Phone Usage (UbiComp ’15)

When Attention is not Scarce – Detecting Boredom from Mobile Phone Usage from Martin Pielot

In times of information overload, attention has become a limiting factor in the way we consume information. Hence, researchers suggested to treat attention as a scarce resource coined the phrase attention economy. Given that attention is also what pays the bills of many free internet services through ads, some even speak of the Attention War. Soon, this war may start extending to our mobile devices, where already today, apps try to engage you through proactive push notifications.

Yet, attention is not always scarce. When being bored, attention is abundant, and people often turn to their phones to kill time. So, wouldn’t it be great if more services sought your attention when you are bored and left you alone when you were busy?

Since mobile phones are often used to kill time, we — that’s Tilman Dinger from the hciLab of the University of Stuttgart, and Jose San Pedro Wandelmer, Nuria Oliver, and me from Telefonica’s scientific group — saw an opportunity in detecting those moments automatically. If phones knew when their users are killing time, maybe they could suggest them to make better use of the moment.

To identify, which usage patterns are indicative for boredom, we logged phone usage patterns of 54 volunteers for 2 weeks. At the same time, we asked them to frequently report how bored they felt. We found that patterns around the recency of communication activity, context, demographics, and phone usage intensity were related to boredom.

These patterns allow us to create a model that predicts when a person is more bored than usual with an AUCROC of 74.5%. It achieves a precision of over 62%, when its sensitivity is tuned detecting 50% of the boredom episodes.

While this is far from perfect, we proved its effectiveness in a follow-up study: we created an app (available on Google Play, more info here) that, at random times, created notifications, which suggest to read news articles.

When predicted bored, the participants opened those articles in over 20% of the cases and kept reading the article for more than 30 seconds in 15% of the cases. In contrast, when they were not bored, they opened the article in only 8% of the cases and kept reading it for more than 30 seconds in only 4% of the cases.

Statistical analysis shows that the predicting accounts for significant share of the observed increase.

While we certainly don’t feel that recommending Buzzfeed articles will be the cure peoples’ boredom, at least not for the majority of them, the study provides evidence that the prediction works.

Now how can mobile phones better serve users, when they can detect phases of boredom? We see four application scenarios:

Engage users with relevant contents to mitigate boredom,
Shield users from non-important interruptions when not bored,
Propose useful but not necessarily boredom-curing activities, such as clearing a backlog of To Do’s or revisiting vocabulary lists, and
Suggest to stop killing time with the phone and embrace boredom, as it is essential to creative processes and self-reflection.

Relatedly to this work, in a follow-up study, we also showed that mobile phones can predict the boredom proneness, the predisposition of experiencing boredom.

The work was presented in September 2015 at the ACM International Joint Conference on Pervasive and Ubiquitous Computing, taking place in Osaka, Japan, where it received best-paper award.

When Attention is not Scarce – Detecting Boredom from Mobile Phone Usage.
Martin Pielot, Tilman Dingler, Jose San Pedro, and Nuria Oliver
UbiComp’ 15: ACM International Joint Conference on Pervasive and Ubiquitous Computing, 2015.

Share this:

Boredom-Triggered Proactive Recommendations

Boredom-Triggered Proactive Recommendations from Martin Pielot

The business model of many internet-service companies is primarily build around your attention: they offer best-in-class services for free in exchange for the users’ eyeballs, i.e. them paying attention to the contents of the services they offer. They pay for their expenses and generate revenue by selling the attracted attention to companies and individuals who’d like to promote their content.

One of the upcoming frontiers in this battle for the user’s attention are mobile devices. Engagement is now defined by push-driven notifications rather than the traditional pull-driven experience. Recommendations will become proactive and notifications will be one essential path to deliver them.

In this battle, we may be facing the tragedy of the commons: when individual companies behave rationally according to their self-interest by increasing their attempts to seek people’s attention, they behave contrary to the best interests of the whole group by depleting the attentional resources of the user and risk that people develop notification blindness (as an analogy to banner blindness).

Attention is not always scarce

However, attention is not always scarce. For example, when people are bored, attention is abundant, and people often turn to their phones to kill time.

In our recent research on When Attention is not Scarce – Detecting Boredom from Mobile Phone Usage, presented in September 2015 at the ACM International Joint Conference on Pervasive and Ubiquitous Computing, we showed that it is possible to detect phases of boredom from how people use their mobile phones. As part of the same research project, we showed that people are more likely to engage with suggested content when they are bored, as inferred by the detection algorithm.

Boredom-Triggered Proactive Recommendations

This finding opens the door to using boredom as a content-independent trigger for proactive recommendations. Assuming that proactive recommendations delivered via mobile phone notifications will become more common in the future, using boredom as trigger will benefit service providers as well as the end users:
End users will receive fewer recommendations that are triggered during times when they are busy. Service providers can use it to reduce the fraction of unsuccessful recommendations, which, for example, decreases the likelihood that users develop notification blindness towards proactive recommendations.

The results will be presented at the Workshop: Smarttention, Please! Intelligent Attention Management on Mobile Devices — Workshop @ MobileHCI ’15: ACM International Conference on Human-Computer Interaction with Mobile Devices and Services, 2015 to be held from Aug 24 – 27 in Copenhagen, Denmark.

Boredom-Triggered Proactive Recommendations.
Martin Pielot, Linas Baltrunas, and Nuria Oliver.
Smarttention, Please! Intelligent Attention Management on Mobile Devices — Workshop @ MobileHCI, 2015.

Share this:

Quantifying Attentiveness towards Mobile Messaging (MobileHCI ’15)

Social norm has it that people are expected to respond to mobile phone messages quickly. However, notifications may arrive to our mobile phones at any place and any time, which, depending on the concurrent activity, can be disruptive for the user. Hence, research has explored ways to reduce the chance of disrupting users by deferring the delivery of notifications until opportune moments.

For most people, the majority of notifications come from messengers, such as WhatsApp or SMS. This type of communication goes along with high this type of communication social expectations. The majority of the people expect people with whom they frequently communicate to respond within a few minutes. Thus, deferral cannot be indefinite: it requires a bound, that is, a maximum delay before the notification is delivered, not matter how disrupting it might be.

But, what is the right bound?
Social expectations suggest a few minutes maximum. However, how likely is it that an opportune moment occurs within 5 minutes?

We collected evidence regarding these questions in our work I’ll be there for you: Quantifying Attentiveness towards Mobile Messaging.

boxhour — This diagram visualizes how attentive people where predicted to be on average for the different hours of the day.

boxday — This diagram visualizes how attentive people where predicted to be on average for the different days of the week.

Over the course of two weeks, we collected more than 55,000 message notifications from 42 mobile phone users. On the basis of this data, we trained our previously described machine-learning model to predict attentiveness. This model uses sensor data from the phone to predict with close to 80% accuracy, whether a mobile phone user will attend to a message within 2 minutes or not.

We used this model to compute each participant’s predicted attentiveness for each minute of the study. In summary, our data shows that people are attentive to messages 12.1 hours of the day, attentiveness is higher during the week than on the weekend, and people are more attentive during the evening. When being inattentive, people return to attentive states within 1-5 minutes in the majority (75% quantile) of the cases.

Consequently, a bound of 5 minutes or less will ensure that bounded deferral strategies are likely to deliver messages in opportune moments, while reducing the likelihood to violate social expectations.

The results are presented at MobileHCI ’15: ACM International Conference on Human-Computer Interaction with Mobile Devices and Services.

Tilman Dingler and Martin Pielot
I’ll be there for you: Quantifying Attentiveness towards Mobile Messaging
MobileHCI ’15: ACM International Conference on Human-Computer Interaction with Mobile Devices and Services. 2015.

Share this:

Exporting RandomForest Models to Java Source Code

This post shares a tiny toolkit to export WEKA-generated Random Forest models into light-weight, self-contained Java source code for, e.g., Android.

It came out of my need to include Random Forest models into Android apps.

Previously, I used to use Weka for Android. However, I did not find a way to export a Random Forest model in a way that my apps can load it reliably across devices, so the apps had to compute the model on each start — which can take minutes.

androidrf solves the problem in a simple way: a python script parses the console output of WEKA when training a RandomForest model with the -printTree option enabled. Then, it creates a single Java source file implementing those trees with simple if-then statements.

The library ships with three additional Java classes that allow to run and test the generated classifiers.

The code is available on Github under the MIT Licence: androidrf

How to use it

(for people who are familiar with WEKA):

Load your data set into WEKA, choose RandomForest as classifier, and enabled the ‘printTrees’ option for your RandomForest classifier. Hint: limit the depth of the trees with the ‘maxDepth’ option, because otherwise the resulting source files may become huge.

Save the output of the results buffer into a .txt file. Best, save it into the ‘data’ folder of the androidrf project.

Open a terminal, enter the ‘data’ folder of the androidrf project, and execute
python to_java_source.py -M filename (without .txt).

A class with the name FilenameRandomForest should appear in androidrf/src/org/pielot/rf

All you need to do is to copy the Java class together with the three pre-existing Java classes (Prediction, Evaluation, RandomForest) into your project. It should compile without error.

The features have been added as fields to your classifier. Hence, in order to specify the features, simply populate those fields. Then, run runClassifiers(List predictions) to obtain a Prediction with the details of the prediction (predicted class, certainty, ..).

Voila! You have a light-weight, portable, working Random Forest model.

Share this:

Cost Explosion in the Health System

Research projects on eHealth are often motivated by the so-called cost explosion that we are supposed to face in Europe. For example, The Economist Insights writes that “basic problem is the spiralling cost of healthcare” and that “healthcare systems [..] are facing financial ruin“.

I used to believe this prediction until a very good book called Lügen mit Zahlen: Wie wir mit Statistiken manipuliert werden (Lying with Numbers: How we are being manipulated with statistics) by Gerd Bosbach, Jens Jürgen Korff offered a different perspective. Since the book is only available in German, I decided to re-do their calculations on my own and present the results in English.

I downloaded the total yearly expenses of the German public health system from the official institution: Gesundheitsberichtserstattung des Bundes. This is how the numbers look:

Phew. Certainly not an explosion, but the curve is clearly pointing skywards.

But wait! Did you notice the y-axis? It does not start at 0, a common trick, as pointing out by Bosbach and Korff, to make changes look more dramatic.

Let’s fix the y-axis:

Okay. That looks less worrying now. But still, the curve points upward. The cost may not be “spiraling”, but the numbers are certainly getting bigger.

But wait! Numbers are always getting bigger. This is called inflation. So let’s put these number into perspective by showing them as fraction of the German GDP (source: statista.com):

Look at that! Expenses have been hovering are around 10-11% of the GDP. It seems convincing that a society should spend a stable fraction of its wealth on health.

Admittedly, there is a slight increase. If I draw a linear trend line on this diagram, health cost will be 16.8% of the GDP in 2060 — but how can really tell what will happen in those 45 years from now.

On the other hand, this is also a matter of how to select the data. If I had only shown data from 2009 – 2013, a trend line computed in these figures would even have looked as if the relative health costs were decreasing.

So, next time somebody pull the “cost-explosion-in-health-system” card, hit them with facts.

Share this: