Difference between revisions of "Peer rating"

From Testiwiki
Jump to: navigation, search
(idea of non-linear exchange rate for onor-money)
(old onor rating table removed, instructions added)
Line 1: Line 1:
 
[[Category:Quality control]]
 
[[Category:Quality control]]
 
{{method|moderator=Jouni}}
 
{{method|moderator=Jouni}}
'''Peer rating''' is a [[method]] to give merit ratings or [[onor]]s to a page in [[Opasnet]]. Peer rating has a scale from 0 to 100, but it can be converted to [[onor]]s with an inverse logarithmic transformation (1 on the peer rating scale equals 0.13 onors, 10 equals 1 onor, and 100 equals 1,000,000,000 onors).
+
'''Peer rating''' is a [[method]] to give merit ratings or [[onor]]s to a page in [[Opasnet]]. Peer rating has a scale from 0 (very poor) to 100 (outstanding). Previously this number was converted to onors using a specific conversion function, but later it was understood that it is not clear what the function should be. The rating has two parts: scientific quality, and usefulness. Both properties are opinions of the person who makes the rating. Both evaluate the actual content of the page at the time of evaluation. Note that the average rating shown on the rating bar may be largely based on the historical quality of the page. Methods to compensate for page improvements is under development.
 +
 
 +
'''Practical hints:
 +
* A person needs to login to be able to make peer ratings.
 +
* Internet Explorer contains deliberate bugs (i.e., deviations from the standard html; the bugs are installed by Microsoft) that may cause trouble using and viewing the rating bar. Improvements are under way to overcome the problem, but in the meantime, we recommend that you complain directly to Microsoft.
  
 
==Scope==
 
==Scope==
Line 11: Line 15:
 
==Definition==
 
==Definition==
  
The [[rating bar]] functionality only accepts integers between 0 and 100. Therefore, it is practical to make the first attempt of peer rating reflect this interval. However, because ratings must be able to be used without limits, it might be possible to make an inverse log transformation using the following equation:
+
The [[rating bar]] functionality only accepts integers between 0 and 100. Therefore, it is practical to make the first attempt of peer rating reflect this interval.
 
 
onor = 10^(rating/10) * 0.1
 
 
 
This results in the range from 0.01 to 100,000,000 onors which should be wide enough for practical needs. However, it does NOT make it possible to give negative respect.  
 
  
 
{{discussion
 
{{discussion
Line 23: Line 23:
 
|Argumentation =
 
|Argumentation =
  
{{attack|1 |Should we change this in such a way that rating below 20 would use this equation:
+
{{attack|1 |Instead, we could do so that if you want to give negative rating, you actually should make a peer review instead. The peer review result may be negative as well as positive.
 
 
onor {{=}} -10^((10-rating)/10) * 0.1?
 
 
 
This would only make it possible to give -10 onors which may not be enough if negative onors are given in the first place. Instead, we could do so that if you want to give negative rating, you actually should make a peer review instead. The peer review result may be negative as well as positive.
 
 
|--[[User:Jouni|Jouni]] 12:29, 22 March 2010 (UTC)}}
 
|--[[User:Jouni|Jouni]] 12:29, 22 March 2010 (UTC)}}
 
}}
 
}}
  
 
==Result==
 
==Result==
 +
 +
[http://en.opasnet.org/en-opwiki/index.php?title=Peer_rating&oldid=15158 A previously used onor rating table.]
 +
 +
A special table will be developed to rank pages based on their scientific quality and usefulness. Then the users can compare their estimates to other people's estimates of other pages.
  
 
The total [[onor]]s earned by an individual i are calculated in the following way:
 
The total [[onor]]s earned by an individual i are calculated in the following way:
  
  O<sub>i</sub> = &Sigma;<sub>p</sub> O<sub>p</sub> n<sub>p</sub> f<sub>i,p</sub>,
+
  O<sub>i</sub> = &Sigma;<sub>p</sub> E(O<sub>p</sub>) f<sub>i,p</sub>,
  
where O<sub>i</sub> is the amount of onors earned by individual i, O<sub>p</sub> is the amount of onors per one reading given to a particular page p, n<sub>p</sub> is the number of times the page p was read, and f<sub>i,p</sub> if the fraction of merit that the contributor i earns for his/her work on page p.
+
where O<sub>i</sub> is the amount of onors earned by individual i, E(O<sub>p</sub>) is the expectation of the amount of onors given to a particular page p, and f<sub>i,p</sub> if the fraction of merit that the contributor i earns for his/her work on page p.
  
 
Onors can be exchanged to money, but in a non-linear manner. There is a threshold for exchange, and onors below the threshold are not exchanged. The actual level of threshold is determined by the members of the society. The amount of exchangeable onors is calculated in the following way:
 
Onors can be exchanged to money, but in a non-linear manner. There is a threshold for exchange, and onors below the threshold are not exchanged. The actual level of threshold is determined by the members of the society. The amount of exchangeable onors is calculated in the following way:
  
  O(E)<sub>i</sub> = &Sigma;<sub>p</sub> O<sub>p</sub> n<sub>p</sub> f<sub>i,p</sub> (if O<sub>p</sub> f<sub>i,p</sub> >= T then 1 else 0),
+
  O(E)<sub>i</sub> = &Sigma;<sub>p</sub> O<sub>p</sub> f<sub>i,p</sub> (if O<sub>p</sub> f<sub>i,p</sub> >= T then 1 else 0),
  
 
where T is the threshold. A major reason for this non-linearity is that anyone can do a large number of low-onor edits without specific skills. With a linear exchange rate, most of the money would be exchanged to such low-value contributions. It is more motivating and cost-effective to give higher prizes for hard and skilled work and none for others. This motivates people (who want to do this for living) to work especially on high-value things.
 
where T is the threshold. A major reason for this non-linearity is that anyone can do a large number of low-onor edits without specific skills. With a linear exchange rate, most of the money would be exchanged to such low-value contributions. It is more motivating and cost-effective to give higher prizes for hard and skilled work and none for others. This motivates people (who want to do this for living) to work especially on high-value things.
 
{| width="500px" {{prettytable}}
 
!Peer rating
 
![[Onor]]s
 
!Interpretation, examples
 
|----
 
|| 0|| 0.00|| Not rated
 
|----
 
|| 1|| 0.13|| An attempt to raise an issue.
 
|----
 
|| 2|| 0.16||
 
|----
 
|| 3|| 0.20||
 
|----
 
|| 4|| 0.25||
 
|----
 
|| 5|| 0.32||
 
|----
 
|| 6|| 0.40||
 
|----
 
|| 7|| 0.50||
 
|----
 
|| 8|| 0.63||
 
|----
 
|| 9|| 0.80||
 
|----
 
|| 10|| 1.00|| A placeholder or a page title that makes sense. (Placeholder)
 
|----
 
|| 11|| 1.30||
 
|----
 
|| 12|| 1.60||
 
|----
 
|| 13|| 2.00||
 
|----
 
|| 14|| 2.50||
 
|----
 
|| 15|| 3.20||
 
|----
 
|| 16|| 4.00||
 
|----
 
|| 17|| 5.00||
 
|----
 
|| 18|| 6.30||
 
|----
 
|| 19|| 8.00||
 
|----
 
|| 20|| 10.00|| A brief page with useful links or a summary. (Draft)
 
|----
 
|| 21|| 13.00||
 
|----
 
|| 22|| 16.00||
 
|----
 
|| 23|| 20.00||
 
|----
 
|| 24|| 25.00||
 
|----
 
|| 25|| 32.00||
 
|----
 
|| 26|| 40.00||
 
|----
 
|| 27|| 50.00||
 
|----
 
|| 28|| 63.00||
 
|----
 
|| 29|| 80.00||
 
|----
 
|| 30|| 100.00|| A good description of at least one part of the topic. Links to relevant additional sources. Draft estimates, if a quantitative variable.
 
|----
 
|| 31|| 130.00||
 
|----
 
|| 32|| 160.00|| A news in a newspaper or website.
 
|----
 
|| 33|| 200.00||
 
|----
 
|| 34|| 250.00||
 
|----
 
|| 35|| 320.00||
 
|----
 
|| 36|| 400.00||
 
|----
 
|| 37|| 500.00||
 
|----
 
|| 38|| 630.00||
 
|----
 
|| 39|| 800.00||
 
|----
 
|| 40|| 1000.00|| A good text that helps to get an overview. Far from exhaustive. Fairly good quantitative estimates, if a variable.
 
|----
 
|| 41|| 1300.00||
 
|----
 
|| 42|| 1600.00|| A large newspaper article.
 
|----
 
|| 43|| 2000.00||
 
|----
 
|| 44|| 2500.00||
 
|----
 
|| 45|| 3200.00||
 
|----
 
|| 46|| 4000.00||
 
|----
 
|| 47|| 5000.00||
 
|----
 
|| 48|| 6300.00||
 
|----
 
|| 49|| 8000.00||
 
|----
 
|| 50|| 10000.00|| A good, peer-reviewed scientific article. A good assessment with a coherent set of variables.
 
|----
 
|| 51|| 13000.00||
 
|----
 
|| 52|| 16000.00||
 
|----
 
|| 53|| 20000.00||
 
|----
 
|| 54|| 25000.00||
 
|----
 
|| 55|| 32000.00||
 
|----
 
|| 56|| 40000.00|| A good meta-analysis
 
|----
 
|| 57|| 50000.00||
 
|----
 
|| 58|| 63000.00||
 
|----
 
|| 59|| 80000.00|| A good novel.
 
|----
 
|| 60|| 100000.00|| An outstanding scientific article. Exhaustive description ot the topic.
 
|----
 
|| 61|| 130000.00||
 
|----
 
|| 62|| 160000.00|| A good scientific textbook.
 
|----
 
|| 63|| 200000.00||
 
|----
 
|| 64|| 250000.00||
 
|----
 
|| 65|| 320000.00||
 
|----
 
|| 66|| 400000.00||
 
|----
 
|| 67|| 500000.00||
 
|----
 
|| 68|| 630000.00||
 
|----
 
|| 69|| 800000.00||
 
|----
 
|| 70|| 1000000.00|| A major scientific breakthrough that changes scientific thinking.
 
|----
 
|| 71|| 1300000.00||
 
|----
 
|| 72|| 1600000.00|| A Shakespeare play
 
|----
 
|| 73|| 2000000.00||
 
|----
 
|| 74|| 2500000.00||
 
|----
 
|| 75|| 3200000.00||
 
|----
 
|| 76|| 4000000.00||
 
|----
 
|| 77|| 5000000.00||
 
|----
 
|| 78|| 6300000.00||
 
|----
 
|| 79|| 8000000.00||
 
|----
 
|| 80|| 10000000.00||
 
|----
 
|| 81|| 13000000.00||
 
|----
 
|| 82|| 16000000.00||
 
|----
 
|| 83|| 20000000.00||
 
|----
 
|| 84|| 25000000.00||
 
|----
 
|| 85|| 32000000.00||
 
|----
 
|| 86|| 40000000.00||
 
|----
 
|| 87|| 50000000.00||
 
|----
 
|| 88|| 63000000.00||
 
|----
 
|| 89|| 80000000.00||  F = G m<sub>1</sub> m<sub>2</sub> / r<sup>2</sup>
 
|----
 
|| 90|| 100000000.00|| E = m c<sup>2</sup>
 
|----
 
|| 91|| 130000000.00||
 
|----
 
|| 92|| 160000000.00||
 
|----
 
|| 93|| 200000000.00||
 
|----
 
|| 94|| 250000000.00||
 
|----
 
|| 95|| 320000000.00||
 
|----
 
|| 96|| 400000000.00||
 
|----
 
|| 97|| 500000000.00||
 
|----
 
|| 98|| 630000000.00||
 
|----
 
|| 99|| 800000000.00||
 
|----
 
|| 100|| 1000000000.00||
 
|----
 
|}
 

Revision as of 15:09, 30 August 2010


Peer rating is a method to give merit ratings or onors to a page in Opasnet. Peer rating has a scale from 0 (very poor) to 100 (outstanding). Previously this number was converted to onors using a specific conversion function, but later it was understood that it is not clear what the function should be. The rating has two parts: scientific quality, and usefulness. Both properties are opinions of the person who makes the rating. Both evaluate the actual content of the page at the time of evaluation. Note that the average rating shown on the rating bar may be largely based on the historical quality of the page. Methods to compensate for page improvements is under development.

Practical hints:

  • A person needs to login to be able to make peer ratings.
  • Internet Explorer contains deliberate bugs (i.e., deviations from the standard html; the bugs are installed by Microsoft) that may cause trouble using and viewing the rating bar. Improvements are under way to overcome the problem, but in the meantime, we recommend that you complain directly to Microsoft.

Scope

What is such a method to give respect or merit to contributions (Opasnet pages) such that the method

  • reflects the true social respect on an absolute scale,
  • can be learned and used easily.

Definition

The rating bar functionality only accepts integers between 0 and 100. Therefore, it is practical to make the first attempt of peer rating reflect this interval.

How to read discussions

Statements: Negative respect should also be possible with peer rating.

Resolution: Not accepted. Negative respect must be given through peer review.

(Resolved, i.e., the resolution has been updated to the main page.)

Argumentation:

1 : Instead, we could do so that if you want to give negative rating, you actually should make a peer review instead. The peer review result may be negative as well as positive. --Jouni 12:29, 22 March 2010 (UTC)


Result

A previously used onor rating table.

A special table will be developed to rank pages based on their scientific quality and usefulness. Then the users can compare their estimates to other people's estimates of other pages.

The total onors earned by an individual i are calculated in the following way:

Oi = Σp E(Op) fi,p,

where Oi is the amount of onors earned by individual i, E(Op) is the expectation of the amount of onors given to a particular page p, and fi,p if the fraction of merit that the contributor i earns for his/her work on page p.

Onors can be exchanged to money, but in a non-linear manner. There is a threshold for exchange, and onors below the threshold are not exchanged. The actual level of threshold is determined by the members of the society. The amount of exchangeable onors is calculated in the following way:

O(E)i = Σp Op fi,p (if Op fi,p >= T then 1 else 0),

where T is the threshold. A major reason for this non-linearity is that anyone can do a large number of low-onor edits without specific skills. With a linear exchange rate, most of the money would be exchanged to such low-value contributions. It is more motivating and cost-effective to give higher prizes for hard and skilled work and none for others. This motivates people (who want to do this for living) to work especially on high-value things.