Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Top colleges should select randomly from a pool of ‘good enough’ (2005)

chronicle.com

Post reply on HN
Print view View original

201–210 of 307 posts

Re: Top colleges should select randomly from a pool of ‘good enough’ (2005)

#201

Post by tyingq »

chrisseaton wrote: post #198
tyingq wrote: post #163

Earlier quoted context omitted.

They weren't claiming. Your scores are all posted on the wall in basic training. I walked the dorms and saw ~5 aces or very near aces out of maybe 200ish total people's scores. It's just somewhat odd because starting out as an E1 isn't terribly lucrative. It doesn't surprise me that 5 of 200 do that well. It surprises me that 5 of 200 that do so well actually enlist.

> It surprises me that 5 of 200 that do so well actually enlist. Can't be that surprising that intelligent people value the experience of being in the military?

In the military as bottom of the rung enlisted people, yes. I feel okay saying that, since I was there, and I was one. I had, er, blown some earlier opportunities. It's not an easy life for the first couple of years, and the pay is terrible at first.

Re: Top colleges should select randomly from a pool of ‘good enough’ (2005)

#202

Post by chrisseaton »

tyingq wrote: post #201
chrisseaton wrote: post #198

Earlier quoted context omitted.

> It surprises me that 5 of 200 that do so well actually enlist. Can't be that surprising that intelligent people value the experience of being in the military?

In the military as bottom of the rung enlisted people, yes. I feel okay saying that, since I was there, and I was one. I had, er, blown some earlier opportunities. It's not an easy life for the first couple of years, and the pay is terrible at first.

> In the military as bottom of the rung enlisted people, yes.

Some people just want to get stuck into the work and tactics rather than all the fuss of being an officer! Being a private soldier in many units can be an extremely challenging and rewarding experience. Yes the pay isn't great for people with other opportunities, but money isn't everything in life.

I know many people who could easily commission but want to stay operators because that's what they enjoy.

Re: Top colleges should select randomly from a pool of ‘good enough’ (2005)

#203

Post by lozaning »

crazygringo wrote: post #181

> Should Select Randomly From a Pool of ‘Good Enough’ But... they already do . I can't find links right now, but deans of admissions from Harvard, Yale, Princeton, etc. all talk about how they could admit 10x the students and they'd all be equally good. And that their selections aren't exactly "random" but are focused on diversity: that there are enough players for each instrument in the orchestra, athletes for each…

Of all the kids in my IB class in high school, the kid that played the bassoon got into more schools than any of the rest of us, even with moderately worse grades.

Re: Top colleges should select randomly from a pool of ‘good enough’ (2005)

#204

Post by strikelaserclaw »

duncan_bayne wrote: post #43

Absolutely. Because we want our doctors and engineers to be just "good enough", none of this elitism for us! I'm sure they'll still cure cancer and get us to Mars, right? And the truly elite who miss out by random chance ... that doesn't matter because justice doesn't require us to recognise and appropriately reward genius. In fact, screw those priviliged geniuses, they're only first in line because they won a geneti…

i think the point the article was trying to make is that good enough is a much more quantifiable metric and randomization after that is better at minimizing bias from those who select who is "special." I think of it more like, everyone with a 1500 sat go into a hat, and we pick 20 randomly, rather than lets put everyone with a sat score of 1000 in a hat because that is "good enough."

Re: Top colleges should select randomly from a pool of ‘good enough’ (2005)

#205

Post by jacobwilliamroy »

hardwaregeek wrote: post #130

One interesting example is the UT system. Because affirmative action was banned in Hopwood v. Texas, the UT system resorted to a program where they accept the top 10% of each high school's class into all UT schools. It's an interesting way of doing admissions, albeit not perfect. One critique that I'm anticipating seeing is that this encourages going to a mediocre high school. Except...getting good students to go to…

Bad schools are made of bad policy, not bad kids.

Re: Top colleges should select randomly from a pool of ‘good enough’ (2005)

#206

Post by JoshuaDavid »

One point about this that I think is overlooked is that if you don't include some sort of randomness, you'll end up over-optimizing for whatever you're measuring, and under-optimizing for the thing you actually care about. More concretely, if we say that there are some underlying factors which correlate to both student "quality" and to admissions scores, and some random factors, we'll find that score and quality are correlated, but the highest quality students and the ones with the highest scores aren't the same ones. This is easier to explain using a graph: if we assume that "score" and "quality" are two gaussian variables, we get a graph that looks like

    16 ^ Quality
    15 |                 #####BBBB  
    14 |              ########BBBBA 
    13 |           ###########BBBBAA
    12 |         #############BBBBAA
    11 |       ###############CCCCAA
    10 |     #################CCCCAA
     9 |    ##################CCCCA 
     8 |   ###################CCCC  
     7 |  ####################CCC   
     6 | #####################CC    
     5 | #####################      
     4 | ###################        
     3 | #################          
     2 |  #############             
     1 |   #########                
       +----+----+----+----+----+----+>
       0    5    10   15   20   25   30
                    Score
If you do the naive thing, and grab the very highest-scoring students, we set the cutoff at 27, and get all 10 of the students labeled by A in the above diagram, thus have a quality distribution of 10 - 14, with the bulk being around 10 - 13, and the average score being 11.5. If we instead set the cutoff at 23, we get 47 students (labelled B for students above the average of the originally selected ones, and C for the ones below that average) -- if we then pick among them randomly, we'll get a flatter distribution ranging from a low score of 6 to a high score of 15, with an average of 10.98. The distribution of scores looks like

    Score | Strict | Cutoff/
          | Cutoff | Random
    ------+--------+--------
       15 |  0.00% |  8.51%
       14 | 10.00% | 10.64%
       13 | 20.00% | 12.77%
       12 | 20.00% | 12.77%
       11 | 20.00% | 12.77%
       10 | 20.00% | 12.77%
        9 | 10.00% | 10.64%
        8 |  0.00% |  8.51%
        7 |  0.00% |  6.38%
        6 |  0.00% |  4.26%
So if you care a lot about simply not getting bad students, and not very much about getting the very best ones, you should do a strict cutoff, since that will give you the highest possible average. If you care more about getting the very best students, and are ok if you let in a few relative duds, you'll do better to lower your cutoff and select randomly from everyone who passed your cutoff.

Elite colleges get their reputation from having students who go on to do great things, so I'd argue that at least for them, the "cutoff plus random selection" makes sense.

Re: Top colleges should select randomly from a pool of ‘good enough’ (2005)

#207

Post by buerkle »

hardwaregeek wrote: post #130

One interesting example is the UT system. Because affirmative action was banned in Hopwood v. Texas, the UT system resorted to a program where they accept the top 10% of each high school's class into all UT schools. It's an interesting way of doing admissions, albeit not perfect. One critique that I'm anticipating seeing is that this encourages going to a mediocre high school. Except...getting good students to go to…

UT Austin is allowed to accept only the top 6%. It also creates weird incentives for students. A friend of mine works in the front office of a local high school and she knows a lot of students who have dropped electives such as band because it lowers their grade point. In Austin area schools AP courses are weighted on a 6.0 scale which makes the problem worse if you want broaden your education and take non-weighted classes.

Re: Top colleges should select randomly from a pool of ‘good enough’ (2005)

#208

Post by tyingq »

chrisseaton wrote: post #202
tyingq wrote: post #201

Earlier quoted context omitted.

In the military as bottom of the rung enlisted people, yes. I feel okay saying that, since I was there, and I was one. I had, er, blown some earlier opportunities. It's not an easy life for the first couple of years, and the pay is terrible at first.

> In the military as bottom of the rung enlisted people, yes. Some people just want to get stuck into the work and tactics rather than all the fuss of being an officer! Being a private soldier in many units can be an extremely challenging and rewarding experience. Yes the pay isn't great for people with other opportunities, but money isn't everything in life. I know many people who could easily commission but want to…

I mean bottom of the rung, E1. You can do 2 years of community college and enter a couple of ranks higher.

Re: Top colleges should select randomly from a pool of ‘good enough’ (2005)

#209

Post by chrisseaton »

tyingq wrote: post #208
chrisseaton wrote: post #202

Earlier quoted context omitted.

> In the military as bottom of the rung enlisted people, yes. Some people just want to get stuck into the work and tactics rather than all the fuss of being an officer! Being a private soldier in many units can be an extremely challenging and rewarding experience. Yes the pay isn't great for people with other opportunities, but money isn't everything in life. I know many people who could easily commission but want to…

I mean bottom of the rung, E1. You can do 2 years of community college and enter a couple of ranks higher.

But again, some people want to be doing not leading! And not everyone wants or is suited to go to college!

Re: Top colleges should select randomly from a pool of ‘good enough’ (2005)

#210

Post by atlgator »

buerkle wrote: post #207
hardwaregeek wrote: post #130

One interesting example is the UT system. Because affirmative action was banned in Hopwood v. Texas, the UT system resorted to a program where they accept the top 10% of each high school's class into all UT schools. It's an interesting way of doing admissions, albeit not perfect. One critique that I'm anticipating seeing is that this encourages going to a mediocre high school. Except...getting good students to go to…

UT Austin is allowed to accept only the top 6%. It also creates weird incentives for students. A friend of mine works in the front office of a local high school and she knows a lot of students who have dropped electives such as band because it lowers their grade point. In Austin area schools AP courses are weighted on a 6.0 scale which makes the problem worse if you want broaden your education and take non-weighted c…

That is an incentive regardless of UT's policy. The top 10 are always fighting to be number 1 or 2 and weighted classes make all the difference.
Post reply on HN