Fork me on GitHub
Math for the people, by the people.

User login

Ratings - new features implemented

Primary tabs

Ratings - new features implemented

I have just deployed the most recent ratings module. The new features include the following:

1. 'Help' link next to each category. If you click on this, a new popup window shows which should contain what is meant by 1, 2, 3... in each of categories.
2. Feature of comment. When you file a rating, you can put free text comment.
3. When you rate object, there is a new Ratings Detail. It contains table with a list of all ratings. You can basically see whole history (with comments).
4. I added date attribute for ratings. It records time the rating is being filed. For old ratings I put 'date unknown' string.
5. There is a new option 'No vote'. If you file a vote, and want to cancel it, just choose 'No vote' instead of number representing your vote.

I hope those features will make the rating system more usable. Especially that now you can affix a comment explaining your votes. Additionally, the list of votes makes the system more clear - e.g. if you vote, you have to be aware of that your actions are actually displayed to others :)

BTW. When you browse through the ratings, please bear in mind that the anti-spamming algorithm is deployed (I know I should not use the spamming word for this!). The details can be found here: http://planetmath.org/?op=getmsg&id=19304

Thanks!
Pawel


There is no bug :) It is simply the results of de-spamming algorithm. As that only rating was made by user classified into clique, its value is severely dropped. If you rate this entry, the value of your rating will count as true rating, and you will increase the average.

We do have few options. First, I can easily turn off showing old ratings, but then the system is not clear enough in my opinion. Second, we can throw away all the past ratings, and hope for people to start playing fair. What do you think?

Pawel

What does "user classified into clique" exactly means in this situation? How do you decide who fits this category?

If that was the *only* rating, why does the entry say that there were 4 votes?

And if that was the only rating, how did a score of 4,5,5,5 become 1,1,1,1?? It doesn't make any sense to me. Exactly how was the score modified?

Alvaro

> Speaking only for myself, I don't mind. If I seem at all
> embarrassed being called upon to explain my old ratings it
> would only be because I might not remember my exact thought
> process in its entirety. But from a moral point of view,
> there is nothing that I would be ashamed of. I never rated
> anything without reading it, out of pure spite. And if I
> rerate anything higher than I originally did, it would only
> be 1 point more, that is, a subtle shift rather than a
> hypocritical change.

My only point here is that one cannot release to the public information that was previously supposed to be private.

I wish everyone wrote reviews from a completely objective point of view. But that's simply not the case. You just have to look at the messages in the forum, all that nasty bickering, to foresee what will happen when some members receive a bad review - with a name attached to it - ... bad ratings will be thrown around, rendering the whole system useless once again.

> Also, there is something broken with the ratings system. See for
> example the entry "circle":
>
> http://planetmath.org/?op=getobj;from=objects;id=4236
>
> The score is 1 across the board (and it seems unlikely that that's > the real score), but there are supposedly 4 ratings. Only one
> rating appears in the history however, and the scores are 4,5,5,5. > Other entries seem to be suffering the same problems.
>
> Alvaro

I am saddened to read about this. It would be in PM's best interests that entries on highly visible topics like "circle" not only be good and correct but also have ratings that inspire confidence in outsiders brought here by Google.

It looks as if "circle" has very little in it written by primefan yet his owning must've been enough for it to become a lightning rod for knee-jerk negative ratings.

It's not terribly bad, though it could use sprucing up. Like... a picture of a circle, for starters! Still, the 1 across the board rating is excessive, and wrong. Overall 3 would make more sense, but more importantly, people need to start working on it to make it good.

Because it's things like "circle" that you can really show to outsiders. Things like category theory and base 10 Angelini numbers are nice to have just for the sake of completeness, but, honestly, they don't do anything for PR.

Jamie

I've thought about that too, but drawing a simple circle can be exceedingly difficult here. It might be possible to do it in pstricks. It can certainly be done in Photoshop, saved in the preferred file format and uploaded here. Every available option has its fair share of downsides.

1and2and4 wrote:
"It looks as if "circle" has very little in it written by primefan yet his owning must've been enough for it to become a lightning rod for knee-jerk negative ratings."

The "knee-jerk negative ratings" are the product of a computer program -- the so called "de-spamming algorithm." The problem that it is intended to address is that of spurious high ratings. When the algorithm encounters an entry that it classifies as highly rated because of collusion, that entry's rating is changed to 1 across the board. Personally, I don't think this is an optimal solution; but I would like to dispel the notion that the low ratings were somehow due to personal animosity.

ratboy wrote:
"Personally, I don't think this is an optimal solution; but I would like to dispel the notion that the low ratings were somehow due to personal animosity."

What do you make of the example alozano brought forward, "circle"? One individual user rated the entry 1 across the board. At least that entry has had a ton of corrections, so 1 can be justified somewhat.

But I bet alozano can also point examples of individuals rating something all 1 when there are no corrections, no comments, no forum posts, nothing whatsoever. I find it hard to blame a computer program in those cases.

At the end of the day, do ratings help increase the prestige of PM at all? Will anyone consult PM first before they consult Mathworld, arxiv, the OEIS, or even Wikipedia?

It is unfortunate that primefan is the center of so much of the contraversy, but he is. Sorry, primefan.

Here are the historical ratings of some of primefan's entries:

a prime occurs in the Euclid-Mullin sequence no more than once

PrimeFan date unknown 5 5 5 3
CompositeFan date unknown 5 5 5 5
Mravinci date unknown 5 5 5 5
CyclotomicQ date unknown 5 5 5 5
PrimeFan date unknown 5 5 5 5

proof that there are infinitely many primes using the Mersenne primes

CompositeFan date unknown 5 4 5 5
knodeltheory date unknown 5 2 5 5
PrimeFan date unknown 5 5 5 5
CyclotomicQ date unknown 5 5 5 5
CompositeFan date unknown 5 4 5 5
ratboy date unknown 2 3 1 1

least prime factor

PrimeFan date unknown 4 5 5 5
Wkbj79 date unknown 3 5 4 2
Robert_Happelberg date unknown 5 5 5 5
knodeltheory date unknown 5 5 5 5
nsmith2001 date unknown 5 5 5 5
PrimeFan date unknown 4 5 5 5
Mravinci date unknown 5 5 5 5
MathNerd date unknown 5 5 5 5
PrimeFan date unknown 5 5 5 5
AL6390 date unknown 5 5 5 5

Polish notation

CompositeFan date unknown 5 5 5 5
CWoo date unknown 1 1 1 1
Robert_Happelberg date unknown 5 5 5 5
Mravinci date unknown 5 5 5 5
CyclotomicQ date unknown 5 5 5 5

These entries were rated very high, and largely by the same people. These are the sort of entries that the rating program changed to 1 across the board. Prior to Pawel's software revisions, nearly all of primefan's entries were rated very highly.

I share your opinion of the PM ratings -- I think the site would be better without them.

Just my 5 centes. With the de-spamming algorithm, all those ratings by the same group of people (people from cliques) do not count... And with the feature of the ratings listing you can catch such groups, so no one can feel anonymous. We were hoping that those two mechanism will prevent users from trying to exploiting the ratings, but rather will encourage using them in good way.

Pawel

> What do you make of the example alozano brought forward,
> "circle"? One individual user rated the entry 1 across the
> board. At least that entry has had a ton of corrections, so
> 1 can be justified somewhat.

Please read my post more carefully, I never said one individual user rated the entry with 1's across the board. The *system* decided to give a 1's rating across the board, even though only one rating existed (which was all high values).

I think the problem with the current system is that a high score was converted to a low score (because "the user was in the same clique as the author" whatever that means). Instead, a score provided by a user "in the same clique" should have a low weight, compared to users not "in clique". So, if only one score exists, that should be the score, but if two scores exist, the scores by people in the same 'clique' should have a lower weight. Even if the scores of an entry are high, and they are all by people in the same "clique", that doesn't mean all scores should be reduced to 1's. The entry may still be perfectly valid. That's why the system should only change the weight of the votes and not their actual value.

Alvaro

Not anymore. Please take a look at the objects right now. I have fixed this bug If the only rating is rating from user within clique, the rating system acts as if there was no rating at all.

Pawel

> Please read my post more carefully, I never said one individual user rated the entry with 1's across the board. The *system* decided to give a 1's rating across the board, even though only one rating existed (which was all high values).

You did give an example where CWoo's negative vote was enough to knock down the whole score, but I guess it was't circle.

This must mean that the clique was overcompensating for CWoo's attack.

What they must've not realized is that CWoo must be part of the house.

The house always wins.

This is just a test. Please ignore it.

This is also just a test. Please ignore.

Sorry, couldn't help it.

Cam

This is a test post. Please ignore.

mathcam wrote:
> Sorry, couldn't help it.
>
> Cam
>
>
>
Cool - you helped to test the email bridge :) Thanks!! I knew how to
catch someone's attention ;)

Pawel

Hello,

First, the ratings are going to be *public* from now on? and retroactively as well? This seems a very bad idea to me. It will only generate problems, such as vindictive ratings among users. Why not anonymous as they were before?

Also, there is something broken with the ratings system. See for example the entry "circle":

http://planetmath.org/?op=getobj;from=objects;id=4236

The score is 1 across the board (and it seems unlikely that that's the real score), but there are supposedly 4 ratings. Only one rating appears in the history however, and the scores are 4,5,5,5. Other entries seem to be suffering the same problems.

Alvaro

Subscribe to Comments for "Ratings - new features implemented"