Ask HN: A solution for "How to recruit?"
LinkedIn fails due to the eBay problem. You can't publish the connection graph or text-based recommendations and maintain any sense of reporting balance. The results are always biased and too many people are attempting to game the system.
My proposal relies on a system similar to page rank as calculated over a set of hierarchically arranged skills. The more highly regarded the reviewer's related skills, the more certain the system is that the reviewer is a real person, and the more similar their opinions to yours, the more weight their opinion carries. Because it's based on skills, searching for contacts is easy; you simply select the requisite skills in order of importance and can determine a minimal value. For instance, you can look for a software engineer that is skilled in PHP, PostgreSQL, and French whose French is at least fluent.
There are a number of analytics approaches I'm considering to limit the ability of individuals to game the system. Further, the system seeks to incorporate a "humanness" factor. That is, each user's identity is verified as much as possible. For instance, by validating a credit card (and checking it's name against the user's name); the charge on the card would be used to send a certified, restricted-delivery letter via USPS with a confirmation code.
The data would use the ZKDB (zero-knowledge database) design to ensure user privacy and security and no data associated with a user and their opinion of another would be stored on our system. This should allow the user to be confident in their assessment of an individual as being private while allowing the system to rate the opinions of "guest" accounts very weakly. Businesses would be similarly identified.
No reporting would be available for an individual until a certain criteria has been met (number of reviews, et cetera). Thus a user can't easily track changes or attempt to deduce who provided what review to the system. This would help alleviate the eBay problem wherein I'm above average, you're above average, and we're all above average.
Sophisticated analytics would be employed to identify mathematical oddities which indicate gaming attempts like disjoint sets and orbits. To be responsive, the analysis would be done in a distributed fashion on a cloud and then stored in an easily-referenced fashion. The incentive for people to add information is basic quid-pro-quo. You all want an easier way to hire people; the cost is simple, you must evaluate coworkers and employees. Because the system hinges upon aggregate opinions of people and since no one person or group can significantly impact the reputation of an individual, there should be no legal concerns with libel. We're merely rating an individual's skills on a relative scale according to the analytic system of our software. It's the same for everyone and thus cannot target an individual.
12 comments
[ 4.5 ms ] story [ 34.3 ms ] threadI just can't get behind a service that figures out how to solve everyone's recruiting issues by assigning them a Better Number.
He doesn't. But, of course, he doesn't control his ranking -- others do. So, deleting his account doesn't do anything to remove his ranking. Perhaps people who are below average will then drop off and the cultural relevance of a rating of 5 out of 10 will adjust itself. I'm betting that enough will stay to make the site worth while -- certainly anyone for whom the system works. Also... not all companies want a 10-point programmer. Those cost money. Some tasks only require a 5-point developer.
Where I DO see some interesting numbers having values in this area are where the numbers aren't rankings, but rather indications of participation, passion, and relevance. For instance, I think Stack Overflow scores.. or maybe even HN karma points, might both be really valuable indicators. If I got a resume from someone with a Stack Overflow score attached to it, I'd know that this is a person who participates, is a good communicator, and knows the answers to lots of tough questions! Not only that, but I can go look at his or her posts.
I think this is where your project needs attention... What's in it for the employee? If only '10 point' employees really like your service, then I think you're wasting your time.. '10 point' employees don't have problems finding work.
Further, "what's in it for the employee" doesn't enter into it. I could easily say: famousactress is rated as 8.34 for her dramatic performances but only at 5.21 in comedic roles. That has nothing to do with your willing participation. I agree that it's depressing to be considered sub-par. But, there are so many people trying desperately to stand out that they're losing their way.
Google can't hire the right sort of people because every metric they have attempted to date has resulted in honest, dependable, clean-cut, recently-shaven, knowledgeable CS graduates from top universities. Sadly, they're trying to expand a business built (like so many other startups) by the ideas of college drop-outs (or similar), who were too busy doing something interesting to go to class and get good grades. Those interesting things aren't bragging rights (level 80 character in each class and each faction on WoW appears on very few resumes).
If I attach my HN information, you might disagree with my positions and opinions. I'd miss out on the job based upon personal bias -- maybe I should. But, you'll notice my karma here is _very_ low. I suppose that seems justified since you're not on board with this concept. But, is it? Reddit is commonly gamed. If I had an outrageous karma score, would you think highly of me or consider me a karma whore?
Well, you might say, you'd simply read my posts. Great! Now, consider you have a highly coveted position and 10,000 applicants. You can only read the posts of about 20 applicants. Do you take the top 20 highest scores? What if the 21st highest was actually the best fit? Well, then you blame your metric. The reason you blame your metric is because people will game their karma scores to reduce their competition to only 20 other people who were busily whoring karma and not building necessary job skills. At the same time, you can't deal with 10,000 resumes. You need a metric that's not easily gamed and that's as representative as possible of an individual's skills.
Just like I can tag you in a photo on Facebook without your consent, this site would allow me to rate your skills as an actress and as a programmer. My opinion of your acting skill is pretty much worthless (whether it's good or bad); my opinion of your programming skill is more significant -- but I'm not going to single-handedly make you or break you.
Think about the newest social games. They metric is progression. There are no downers. If you have a low score, it's because you're new to the game or haven't played enough. You always "gain" experience points, gold, whatever.
In order to get past the chicken and egg problem in the social space you need to turn the scale from 1 to 10 into 1 to infinity and give the perception of not punishing low scores.
While the perception of a 1 to infinity scale reduces the immediate negativity of being obviously below average... that illusion only works for a short period of time. When you start playing WoW 10 gold coins is a lot of money, you can delude yourself and harbor odd ideas about your worth. However, once you compare yourself to _anyone_ else, it becomes immediately apparent that 10 gold coins is practically worthless. As far as the "misdirection" is concerned, perhaps I ought to note some aspects of the UI.
Unlike LinkedIn or Facebook, the site doesn't revolve around the user. As such, when you visit the site or login, you are not presented with your profile. Rather, the site is devoted to finding and comparing applicants. As such, the main page is a search page and most users will likely spend much of their time sorting through candidates. In fact, the only way a user could see their own data is to search for themselves. My guess is that that'll be a fairly common practice. My hope is that people will identify their weaknesses and either improve them or re-evaluate their career goals.
I know it seems harsh. But, the hiring practice is harsh. I sit people in a little, unfamiliar, echo chamber and send in a string of people to scrutinize and judge them. I'm not trying to get everyone hired on at obscene salaries and stroke everyone's ego. My goal is not to paint the world with a rainbow facade of unicorns and fairies. I want to enable Harvard, Yale, MIT, IBM, Microsoft, Google, et cetera to have a tool which quickly and accurately separates the wheat from the chaff. If an individual's skills are average or worse, they're not likely to be working at a company which receives so many applications that so sort of automated elimination is requisite. Rather, they're probably looking for average people for average work at an average company. And there's nothing wrong with that.
Here on HN, there are a lot of startups with a different sort of problem. No applications, but strong desires/needs for well-qualified employees. In that case, this tool would also be useful. If a startup wants to rely on friends from college and guys met at HN meetups, that's fine.
Again, I appreciate any and all feedback and am aware of the cold start and related problems. I have a few solutions that I'm not interested in discussing here and now; but, I understand that this is a major concern and certainly appreciate people pointing it out.
But, I've got a few ideas about how to do some initial data mining to build a useful-yet-niched database. I'll target a specific industry and, I'm guessing, enough people will value their skills that they'll be happy to encourage their friends and colleges to rate them. The hope is that this can be done quickly enough to generate sufficiently many data points for accurate analysis without the initial dataset.
Once that happens, I'm banking on the utility of the system. If it's as good as I hope, it'll be adopted by other groups. If it's not, then it should fail.
As far as the grossness factor and potential for abuse goes; I'm protecting user information as much as is humanly possible -- at least, to the limits of my own ability. The system will be open sourced and as transparent as possible. Unfortunately, I can only guarantee so much. I can't think of a way for users to verify what code is actually running on the server -- but, they can ensure that the data they are sending to the server is exactly what they expect. But, there is no certainty that the server isn't recording their IP and associating it with the input for future data analysis. Hm.
"Downer" and "demotivational" seem to indicate that you're approaching this from the perspective of an employee. The site does not operate according to the willingness of employees to be evaluated. I am happy to rate my coworkers who are good or bad (in my opinion). I'm happy to be rated by them, my employers, and my peers. Not everyone feels the same way. That's fine. And, yes, there _is_ a cold start problem.
While I'm well aware of the issues with building a useful site (and have many ideas to circumvent them) -- I'm more interested in the perception of the utility of the site were it to exist and have a reasonably large data set.
Are these metrics truly indicative of an individual's skills? No. In fact, many applicants might print their application on hot pink A4 hoping to get noticed (and hoping that attention is positive).
Even with this tool, I'm merely reducing your applicant pool -- not selecting employees for you. My previous company used a few personality profile tests that were very effective in identifying people who'd work well at the company and with one another. That's not really the point, though. The point is determining who is a good Java developer and who isn't. Most recruiters I've dealt with know only buzz words and can't differentiate between well-qualified candidates and those who "used Oracle 10g to port BSD to J2EE over XSLT."
If your preference for UI engineers leans one way or another, the system would make appropriate corrections to the recommendations. Rather than getting a UI engineer who needs to work closely with a back-end developer and visual designer -- one who can pretty much only manage CSS and Flash work -- you can get results that include people who are more comfortable with the MVC framework, can handle templating in various languages, hates action script, can sling javascript in their sleep, and has a keen eye for visual design.
I'm not offering a one-stop, all-inclusive solution. I'm merely suggesting a method by which to intelligently and functionally reduce applicant pools to something the HR department can handle.