Introducing LabRdr: An Experimental Offline News App That Knows You

We’re refreshing content recommendations through transparent use of data, higher degree of control by the user, and minimal thoughtful alerts.

Open any news app during your next commute. Assuming you don’t have cell service or wifi (and are not driving), do you get relevant, up-to-date news? The answer will likely depend on whether you remembered to open the app before going offline. If not, you’re likely to see… well, yesterday’s news. You’ll also get everything that organization had published at the time of that last app update, regardless of your interests or how little time you have to spend ferreting out articles you want to read.

With these limitations in mind, the Guardian Mobile Innovation Lab has built and is now launching LabRdr (pronounced ‘lab reader,’ although, this being the internet, we’re also fine with ‘labrador’), an experimental iOS app meant for offline news reading during commute times. You can download it now from the App Store.

What is LabRdr?

With this app, we’re experimenting with making offline news reading easier and more relevant through automatic personalization of article selections, based on signals like your interests or, possibly in the future, your location and what’s being read nearby.

LabRdr’s approach to offline news reading is experimental, and different from existing offline news apps in a few ways: Rather than give you all current stories on every topic, it delivers only a self-contained package of Guardian articles keyed to your interests, twice a day just in time for your commute, at times you can specify.

As you use it, it learns what you like to read and delivers you content keyed to your interests. (We’re setting aside important conversations about filter bubbles for now to learn something about personalization.) In addition, we show you how we use the data you share with us, in an effort to enhance trust through transparency.

Here are some of LabRdr’s major features:

Alerts only twice a day: You’ll get regular alerts for your reading packages twice a day, at times you specify, for your morning and evening commutes. If you opt into them, we may also send you a small number of experimental alerts, keyed to other personalization signals, like your location.

Offline reading: Reading packages are accessible both on and offline. The app automatically updates in the background twice a day.

News you can finish: The length of each “reading package” is based on the duration of your commute, as you specify, based on total character count. You can adjust the settings to get a longer or shorter reading package.

App gets to know your preferences: LabRdr tracks the categories of the news stories you read, then uses your most frequently read topics to recommend articles for your next reading package.

Transparent use of data: View your reading and commute patterns in the Log — the topics you read often as well as where you read. We ask for permission to access your location data through the app, and we want you to see why (although we won’t be looking at or sharing individual users’ locations!). More on this below.

With this app, we’re experimenting with making offline news reading easier and more relevant, through automatic personalizations of your reading package based on signals like your interests or, possibly in the future, your location and what’s being read nearby.

We also hope LabRdr will save a reader time, since it doesn’t require you to pre-select a lot of broad news categories when you start using it, or spend time saving links to articles you want to read later.

Since it’s experimental (like all of the lab’s projects) the app will only be available temporarily. We’ll remove it from the app store after we finish our experimentation, near the end of the year.

What inspired this idea?

The idea for this app came from a common problem described by a lab engineer, Connor Jennings. Every morning, before he entered his usual subway station in Brooklyn, he would quickly open a handful of browser tabs with articles from The Ringer, a sports news site. Once he was underground he’d look through the tabs and hope that he liked the stories he opened at random.

This messy experience got Connor and the lab team thinking a lot about content recommendations, and how we might be able to use information and signals that are already stored on someone’s mobile phone to improve and customize the offline news reading experience.

Connor researched recommendation algorithms exhaustively and we had a lot of conversations about products or features that exist today to support offline reading or content recommendations. For various reasons it seems like the industry hasn’t quite hit on a successful model. Madeline Welsh, then the lab’s associate editor, even explored the problems and possible solutions deeply in her post, How Do We Build a Better Recommendation Experience for Mobile News Readers?

Ultimately we thought that, by building an app that could allow us to try lots of different algorithms and methods to drive content recommendations, we’d start to gain some deep insights about what makes personalization work for news products.

Our goals

We want to be incredibly transparent about how we are using the data we gather about your commute patterns and the things you choose to read. To that end, for each permission we request, and for each piece of data it allows us to track through the app, we also explain how we’ll use it to aim to improve your commute reading. We’ve also created a section — the Log — where you see what data we’ve collected on your use of the app.

The transparency goes two ways: We’ll also be asking users for feedback about what they like about the app, or what they’d like to see done differently. The reactions we gather from our app users will be analyzed and shared with other news organizations and technologists, in the hopes of helping advance the development of smarter, better-personalized mobile news experiences.

What we’re looking to learn

What makes a good content recommendation system for news? A lot of the existing work about content recommendations are around e-commerce and we’re interested in what signals are particularly good for news organizations and news reading.

We’re also looking to gauge readers’ reactions to the utility of having a short package of news defined for them for a set period of time. Without the option to read a full spectrum of articles on many topics, will they feel better informed with those they do read, or have a sense of achievement at completing a few articles in a set?

As with all our experiments, we’ll report on what we learn in follow-up posts after the app has been running for a while and we’ve collected and analyzed data and reactions.

We hope you’ll download and use the app for the coming week or two. We look forward to your reactions, and please comment here if you have any questions or initial feedback.

The Guardian Mobile Innovation Lab operates with the generous support of the John S. and James L. Knight Foundation.