I am officially getting cynical about LinkedIn Endorsements

[jducoeur]

The hot new feature on LinkedIn is "Endorsements". Basically, they show you a list of randomly selected people in your network, and random skills that they claim to have, and ask whether you can vouch for those skills. It's a fairly clever idea in theory.

The problem is, LinkedIn is trying to have it both ways -- to be both a professional resume site *and* a social network. And the result is that most of the endorsements (anecdotally, based on my experience) don't come from co-workers, they come from friends. Granted, I may be unusual in that I have a lot of friends online, but still -- I think 80% of the endorsements I've received are from people I've never worked with professionally.

The result, of course, is that the endorsements bear little resemblance to reality. So far, I have 2 endorsements each for Perl, Java and JavaScript, 1 for C#, and none for Scala or ActionScript. That is almost *exactly* wrong. I'm currently doing exclusively Scala, spent much of the past ten years working mainly in C# and ActionScript, haven't done Java in over a decade (aside from an abortive attempt to use it for CommYou, which convinced me that I can't stand the language any more), and have only spotty Perl experience at best. Indeed, the endorsements seem to mostly have to do with which buzzwords folks recognize -- that would help account for why I haven't gotten any for things like .NET or Multi-threading, which I actually *am* pretty expert in.

It's kind of too bad, since the high concept of endorsements is a good one. But it's a subtle reminder of why Identity is so important, and why *separating* Identities is important. Having a single LinkedIn identity that encompasses both friends and my professional career actually makes it much *less* useful in some respects, because it muddies the data so much...

Comments

[cvirtue.livejournal.com]

I think some folks use it only for professional links; my own metric is if I know someone well enough to cook them dinner or invite them to crash at my place during hard times, I link to them.

For people who link to anyone they've ever met for 5 minutes, it's a garbage in, garbage out scenario for the endorsements.

[jducoeur]

My own metric is a bit weaker than yours, but much stronger than I use for, say, Facebook.

Still, the whole thing is driving home that most people I'm linked to on LinkedIn really don't know what I *do* very well. I really am ruefully amused to have Perl as one of my top endorsed skills, when I consider it one of my weaker ones...

[meranthi.livejournal.com]

Interesting. I've only been getting endorsements from coworkers so far.

[aishabintjamil.livejournal.com]

My LinkedIn situation is even more odd because I have two accounts - one that has my real name and current professional (for loose definitions) connections, and one for the pseudonym I publish fiction under.

The "real" identity has connections to people at my current employer, past employers, a few people I went to college with, people I volunteer with at CERT, and a slew of SCA friends. The writing one has connections to mostly other writers, and a few friends. But in both cases there lots of people on there that I might feel comfortable calling or emailing and saying "Hey, I see your company is hiring for X, could you put me in touch with someone who'll talk to me about that" or "My employer is looking for a subcontractor to do Y and your company does that - who should I talk to?" or "We need to find someone to do Z, and I know your company used ABC, Inc. for that - what did you think of them?".

None of that makes most of these people qualified to assess my skill set.

[ilaine-dcmrn.livejournal.com]

I've got the problem that I really don't want to say very much about what I do, especially regarding our technology, so my linkedin profile is not much use for anything beyond figuring out if I'm really the person you were thinking of.

[mneme]

In theory.

In practice, I think that one can run analysis, if all the data is present, and filter out endorsements from people who endorse indiscriminately.

After all, one can assume that people will filter into two main categories:

1. People who treat endorsement as a reputation thing and won't endorse someone whose skills they cannot personally vouch for.

2. People who treat endorsement as a "friendly" thing and will endorse their friends regardless of their knowledge of the friends' actual skills

People in cat #2 will tend to do it a lot, which should make a program that wants to use endorsements professionally not -that- hard to filter those out.

[umbran.livejournal.com]

Following the endorsement history of every linkedin user is a bit heavy, and would likely have undesirable unintended consequences - you don't know if someone is endorsing frequently because they are friendly, or because they simply know much about many people. And you don't want to discount the cases when those frequent endorsers have real information, either. Basically, you end up with a disincentive to use - if you use it, what you say matters less!

Instead, we should note that Linkedin already has a concept of the type of connections between people (colleagues, friends, classmates, and so on), as well as a history of where individuals have worked. Simply filtering or weighing results based on relationship is apt to get you far more accurate results: If you aren't a colleague, and haven't worked at the same places I have, you're unlikely to know my skills.

[umbran.livejournal.com]

Linkedin even has a concept of *when* you worked with various people, and could weigh results based on that, as well - someone who worked with you last year is apt to know your current skillset better than someone who worked with you two decades ago...

[mneme]

That was actually my first solution.

The problem with that approach is that it dumps people who endorse friends they don't work with into the same bucket as people who endorse friends they -have- worked on projects with (for instance, I have friends in my larp group who -have- worked on Perl projects with me because we wrote perl to help with production) and, for that matter, people who know each other and have worked with one another in the Open Source world (but not in any professional capacity.

[jducoeur]

True. I've worked with more than a few friends -- indeed, brought many of them into companies I was at. (Including umbran...)

[umbran.livejournal.com]

If you are thinking that working in the Open Source world isn't in a professional capacity, the problem isn't with LinkedIn, but with how you consider your work.

The answer to that one is kind of a "Duh!" - you list projects like that as "places" you've worked, and list those individuals as colleagues!

[jducoeur]

Following the endorsement history of every linkedin user is a bit heavy, and would likely have undesirable unintended consequences - you don't know if someone is endorsing frequently because they are friendly, or because they simply know much about many people. And you don't want to discount the cases when those frequent endorsers have real information, either. Basically, you end up with a disincentive to use - if you use it, what you say matters less!

True -- I suspect that a naive version wouldn't work well. But I'd bet that some smart graph analysis could produce a more nuanced version that can statistically separate folks who endorse indiscriminately from those who are more cautious about it. That wouldn't be *easy*, but I doubt it's any harder than the stuff Memento's been doing to catch crime rings. (Which shares some statistical relationships, I think.)

OTOH, while that would probably produce better answers, it wouldn't be nearly as easy to understand -- and I suspect that transparency is a big priority for LinkedIn here. The current model isn't as useful, but it *is* very simple and obvious...

[umbran.livejournal.com]

Yes, and my answer is pretty darned simple - if you don't want to do the math in a black-box way for them, just give the user viewing the endorsements filters. A list of checkboxes to show colleagues, classmates, friends, only same workplace, and so on. Dirt simple, completely transparent.

[cellio]

Social endorsing is a problem, and that's a pity because (legitimate) endorsements fill a valuable niche below recommendations.

I have endorsed people with whom I've not worked, but with whom I've spent enough time in appropriate contexts to know that they have those clues. Sometimes you can tell by the way someone writes or lectures about a topic, or from volunteer or open-source work.

To the casual observer there's no way to tell the informed endorsements from the social ones. Perhaps better tools will emerge for evaluating this data. (Is the set of endorsements someone has given public? I don't just mean for one person, which you can see from the endorsee's page, but all of them.)

[laurion]

I'm seeing a better balance of where my endorsements are coming from. But that may be that I don't connect to all my friends on linkedin. That's what LJ/FB/G+ are for. On Linkedin I only connect to friends and work contacts that I think are relevent to my field in some fashion, to hopefully manage some of the signal/noise if I ever need to use linkedin for the serious purposes it was originally designed for.

[jducoeur]

The thing is, I disagree about what LinkedIn is for. It's basically an electronic Rolodex, and its primary purpose has *always* been finding "friend-of-a-friend" connections at target companies. It tries to do many other things as well, but they know well that that's their sweet spot, and it's still the primary reason why people actually buy memberships.

That being the case, casting a wide net is appropriate. In my case, I've never actually used LinkedIn for that purpose -- I've got a fairly robust network for job-hunting -- but I am periodically a vector for *other* folks to make connections. So while I don't just friend indiscriminately, I do accept connections from anybody I know pretty well, even if it's not a professional connection; you never know who might want to make a professional connection through you...

[greatsword.livejournal.com]

I haven't endorsed anyone, as it's too generic. I have used the old "recommend this person's work at...." feature, as it allowed an actual explanation of why you recommend that person.

About half my endorsements are from people who could not really judge my abilities.

My rule is that I only link to people I can recommend for at least one thing - even if it is running volunteer efforts at events.

For former coworkers, I only link those that I would want to work with again. Former employees, only those I'd hire again.

I'm considering adding this as a caveat on my linked-in profile, as it will cut down on random link requests from recruiters.

[jducoeur]

My rule is that I only link to people I can recommend for at least one thing - even if it is running volunteer efforts at events.

I'm mostly similar. And indeed, I *am* able to endorse some of those folks for some skills that I've watched them use, or at least heard them talk about to have some faith that they know what they're talking about. But I ignore about 80% of the endorsement requests I get from LinkedIn, and I get the impression that many folks don't.

And I hear you about recruiters. I've linked to the three who I decently well trust (because they have proven themselves over the years to be adequately competent and ethical), but I ignore a lot more...

[mindways]

I got endorsed for skills I don't have and have never claimed, by someone with whom I've only worked in in another field. o.O

[jducoeur]

Weird -- wouldn't have occurred to me that someone can *add* skills to your set. That seems horribly inappropriate (and highly subject to abuse)...