Can we guess your age and income, based solely on the apps on your phone? | Hacker News Reader

Can we guess your age and income, based solely on the apps on your phone?

washingtonpost.com

41 pointscpitman85 comments

They need a box for "I've never even heard of this app" that would allow them to better predict I'm an old curmudgeon who is just fine with missing out on the latest trends.

Instead I'm a twenty-something single woman making more than $52k.

'Never heard of this app' isn't actually discernable form the list of installed apps on your phone, which is at the core of the point they're making.

>They need a box for "I've never even heard of this app"

I would answer that for more than 80% of questions. Not sure how would that help.

You are a single lady younger than 32 who makes less than $52,000/year. Wrong, wrong, wrong and, drumroll, wrong.

I got the same thing. I'm pretty sure they didn't account for security researchers who like to download free apps to hunt for vulnerabilities. Go figure.

Same thing here too, for the opposite reason — said no to almost everything. I have four pages of apps installed, but only three were even on that survey.

Yep, the same for me, looks like a random guess.

I got the same answer, mostly true however.

The real problem is it's a very broad answer anyway.

"You are a married lady older than 32 who makes more than $52,000/year You're either a married, middle-aged lady ... or you use your phone like one."

19 year old male. The only apps that I have on their list are YouTube and Google Maps.

Yep, identical result, though I'm male, 32+ and 52k+.

I do actually use a very small handful of the website versions of some of these (eg, I have been to Walmart's website, I subscribe to some tumblr RSS feeds, I (very occasionally) use Facebook, and I listen to stuff on soundcloud every now and again)..

I generally don't install an App unless there's a very compelling reason to. Most of the time, the App version is going to irritate me with notifications I don't care about, and ask for permissions ranging from questionable to obscene (some game needs access to view all contacts, photos and use my microphone? yeah... no thanks).

Well that's interesting - the only difference between you and I was YouTube, and I'm a married man over 32 making 52,000+. (I have YouTube because Google installed it by default on my phone but I disabled it so I didn't count - also, I'm a 23-year old male who makes a lot less than 52,000/year :D )

I got the same result but also have Yelp installed. I'm a facebook user but m.facebook.com is so much better than the native app.

It is a very sad state of affairs when it is faster to use a website than an app.

And better privacy.

YouTube and Google Maps are pretty bad signals for Android users since they come preinstalled on the majority of phones.

"Either you're a single, middle-aged dude... or you use your phone like one."

I guess the latter is true, for their definition of single, middle-aged dude.

I got that too. Lot of people seem to be saying that here.

Same. Only male. Correct result.

The only app i have is Facebook.

They should add a "were we right?" box at the bottom, I bet the results would be pretty bad. (For me, not a single thing they assume is right)

They could train the algorithm that way too.

That's how Hunch worked and it was pretty good.

>You are a single

Sorry

> guy

Ding!

> younger than 32

>who makes more than $52,000/year

One out of two is ... Better than zero! I went back to the drawing board with a similar prediction system after a series of spectacular failures. I expect it's a function of how the model was trained and tested; it seems that the weights given to any specific app may be over fit. But I'm not a mathematician or statistician or anything beyond a voracious learner with an interest in this kind of thing. It's piqued my curiosity, and I'm enjoying the responses you guys are adding here. Actually, it's already made me think about how my model for my pet project may be over fitting...

(Edited to clarify "single guy" into two features)

The only two apps I had in the list were Youtube and Google Maps, both are installed at the factory and it guessed: You are a married guy older than 32 who makes more than $52,000/year

Exact same result. I have five screens of apps on my phone yet those two and Uber are the only apps I have from the quiz.

So basically the answer is no, they can't really tell.

For what it's worth, I got nailed for a single guy in his twenties over $52k/year.

Same result, except apparently I'm a single female. Must be because I don't play Clash of Clans.

Well, looks like we're all single ladies under the age of 32 who make 52k/year

haha yep :~). I'm baffled why it thought I was single because I use soundcloud. I know people who are married/dating who use that app. Why would that be an indicator of relationship status?

Got me 100% right, which is weird because I figured the lack of apps on my phone would be a problem (I have a windows phone so there's a lot missing). Sounds like it wasn't too accurate for the rest of you though.

Strange. I thought it would be bad for me due to lack of apps (personal choice) too and the only thing it got wrong was age - and even that was pretty close.

If you roll the dice enough times, you'll eventually get snake eyes.

Well right, especially since it looks like there's, what, 2^4 possibilities here? 1 person in this thread getting a correct guess seems about right.

Pretty neat. It's always fun to look at data and find interesting correlations. It nailed it's estimate of me and I felt like I was just clicking no for almost everything.

Same here, some of the apps (like Pandora) don't even work in my country.

"You are a married guy older than 32 who makes more than $52,000/year"

Not really no, granted I ticked no to all the boxes apart from youtube/maps that comes with android

Apparently not using proprietary bloatware makes me a married lady older than 32 that makes >52k$. Well done, Washington Post. 0/4.

"You are a married lady older than 32 who makes more than $52,000/year

You're either a married, middle-aged lady ... or you use your phone like one."

I am a male (strike 1), 50+ (strike 2), Side business takes in more than $52k (strike 3).

I am married (Zillow for Mrs. Mar) and have 2 teen aged daughters (Instagram to follow them).

I'll give machine learning another decade /s.

I see no evidence that they used machine learning. If they had, this thread might be less full of people saying it did worse than random.

From the paper linked in TFA, scikit-learn's logistic regression model (a poorly named one, since it's used for classification rather than regression) is used as the classification model. Like with most machine learning endeavors, I've found that Johnny Five said it best: "More input!"

Based on the fact that the only app I have from the set is Google Maps, we're in the same group.

Except I'm not married, and not a lady.

Facebook has decided I'm a hispanic teenager based on my user profile.

And then the industry wonders why ad blockers are exploding.

Huh? It got everything except your gender correct.

Missing the age by a generation and the income by multiples is not 'correct' IMHO.

> There are 16 possible results, based on your gender (male/female), your age (over/under 32), your marital status (married/single) and your income (over/under $52,000).

They weren't testing at that level of granularity.

They only said they would guess over/under for each of those thresholds, so no, it was correct.

True. I skipped that. I'll pay more attention to the text going forward.

The 61% and 81% accuracy is funny - try to run targeting with that!

The biggest issue with this problem is the data imbalance - how many females over 50 use apps in general compared to 20year males for example. The training data is heavily skewed, which leads to the algo learning that everyone is a single lady under 32 who make 52k/year.

You are a single lady younger than 32 who makes more than $52,000/year

Wrong, wrong, correct!, wrong.

As I suspected, Etsy throws them :) Remove that and they figure out I'm a guy. I guess they can't tell the difference between a single lady buying herself things on Etsy and a guy buying his girlfriend something on there, so, no they can't.

> if you have Uber, you’re probably single.

Really? That's the indication? Not Tinder? You sure Uber doesn't just mean I live in a city? Really sure?

The guess was right though. That was kind of nifty, but my answers would change if the question wasn't do I have the app, but do I use it.

    You are a married lady older than 32 
    who makes more than $52,000/year

the answer you get if you mark no's for all except the two that came with my nexus 5: gmaps, yutoube;

where are my favourite apps on the list?: fdroid, kidsdoodle;

Slightly off with my age but I'm guessing that's because I've recently deleted all but the essential apps from my phone. I'm sure if I still had Snapchat on my phone they could've predicted age better for example.

> You are a married lady older than 32 who makes more than $52,000/year

Good to know that married ladies have older than 32 have squeakymail and keychain on their phones. See - pgp is not that hard after all.

"You are a married lady older than 32 who makes more than $52,000/year"

When actually I am a co-habiting (good as married tbh) man who is 31 and makes £120,000/year. So not the greatest guess hah.

Hey I'm actually moving from the US to the UK and finding the salaries are much smaller. Do you work in tech? How did you get that salary?

I am a C++ programmer (mostly C++ anyway, I also use C and Java albeit not much which I always say I am a C++ programmer as 90% of what I do is C++).

My salary is quite high for a C++ programmer though but that is because I also have responsibility for a lot of things that would normally fall into the technical project or programme manager role. I help with pre-sales, technical support, architecture design and deployment, operational performance tuning, etc.

Also I live in West London where salaries are generally a bit higher anyway. I mostly work from home with a trip to the office in Vauxhall every two weeks during normal operations. The office is only a 30-40 minute train ride for me though and when we have a lot going on I will be in the office a couple of times a week if needed.

My previous experience and specialisation also help me command a higher salary, I have worked for a few large and high profile companies which does give me an edge as such.

Where are you moving to in the UK? What will you be doing?

Looking right now at work in Edinburgh or Glasgow. Sounds like I'll need to make a move to London if I really want to earn the big salary.

I work primarily in native mobile. I'm working for Amazon right now and I've worked for banks, government as well.

Yeah you won't get such high salaries up in Scotland. Outside of London I know Bristol pays well also. I think Manchester has some good things going for it also but never worked there myself.

What exactly do you mean by native mobile? To me that could mean mobile device firmware or Android NDK type thing.

By native mobile I mean writing Java/objc using the mobile SDKs versus something like Cordova or phonegap. Though I have written C++ for The Sims 3 on iOS.

They had me pegged as a married woman over 32 and earning more than $52k. Apart from being male, the rest is correct but there weren't many apps I had and fewer that I used (have Pinterest and Yelp but use neither).

I said yes to ESPN (though I probably have other sports apps instead). Surprised that didn't outweigh Pinterest as a gender marker?

Lookout: yes = guy, no = girl Linkedin: yes = more than $52/K year, no = less than $52K/yr. Snapchat: yes= single,no=married

At least with my base answers.

It got me correct, but not my income. I use the Yelp website, never their app :) Goes for a few of them in their list, as well.

This app made my day, it guessed that I was half my age. The income was way off as well.

But sometimes though I do feel that I am half my age!

I'm a married middle-aged man. Except single and 10 years younger. HN may be an exception; we are odd.

You are a single guy younger than 32 who makes more than $52,000/year

Ding, except I'm in a long term committed relationship. Interesting that it thought I was single (I said no to every single dating/chat app on the list), no facebook, etc, only thing I have is instagram.

"You are a single guy younger than 32 who makes more than $52,000/year".

Not single. Older than 32. Oh well!

Hilariously bad.

"You are a married lady younger than 32 who makes less than $52,000/year"

Wrong on everything.

Every fact about you is normally a tell and sometimes just random: the apps you use, the clothes you dress, the people you know, the travels you made, etc. A fine-tuned algorithm has an high probability to guess this.

Reading from the comments - especially the 52k, they should at least add a geo location.

For m/f the phone name may help.

I understand this is out of scope, but the results seem pretty poor otherwise.

What about the parent who has a zillion apps courtesy of their child? so some apps are the adult's and others are the kids.

Bet that would muddy any statistical analysis.

I wonder if an easier way would be through accessing your camera roll and determining age and gender through the whichever face appears the most.

"You are a married guy older than 32 who makes more than $52,000/year", I am a guy, so... 1 out of 4 I guess.

Haha, iFunny makes you younger than 32.

(Your only other apps being Google Maps and YouTube makes you a married male making 52k+)

If I'm not mistaken, iOS apps aren't allowed to see what other apps are installed, is that correct?

It's nuanced, but you can use techniques like app URL scheme querying to determine what is installed. [1] People will be replying to you with information from 2011—today, it is generally not possible to query for a large number of installed apps this way. So generally, the answer is no.

Without going into much detail, the state of the art uses Facebook in a peculiar way to determine which _relevant_ apps are installed, relevant meaning high grossing (Clash of Clans, Spotify, Tinder are examples).

[1] http://useyourloaf.com/blog/querying-url-schemes-with-canope...

They're allowed to ask "does anything on this system handle yelp:// URLs?".

Nailed me but goofed on gender!

Most Washington Post readers are older. But it would be cheating to use that info.

> You are a single guy younger than 32 who makes more than $52,000/year

Neat.100% right.

After this quiz I definitely feel old: never heard of like 28 out of 32 apps

Apparently I'm a single lady under 32 (wrong on both counts).

Haha I'm married (wrong), over 32 (wrong by about ten years), and make more than 52,000/year (wrong by more than I'd like to admit as an intern :D)

Odd, no choices for ConnectBot or a full Debian install?

I have a stock Cyanogenmod phone with Orbot and Signal. I don't otherwise install apps (less attack surface), so I hit "No" to most of everything on the list.

Amusingly, Betteridge's law of headlines holds true here as well.

This might as well be a joke quiz you share results of on Facebook it's so wildly inaccurate.

completely wrong

Reply on news.ycombinator.com