| Pages in topic: < [1 2 3] > | Non-transparent contest rules and illogical winner determination Thread poster: Zea_Mays
|
|---|
Samuel Murray 
Netherlands
Local time: 20:15
Member (2006)
English to Afrikaans
+ ...
| Okay, I'll feed the troll | Oct 22, 2023 |
Zea_Mays wrote:
I bet you've never heard of giving AI bots a role when writing a prompt, have you?
When you give an AI bot a role, e.g. a lawyer, it doesn't turn the bot into a lawyer (which gives actual legal opinions) -- it just makes the bot sound like a lawyer or makes it give an answer that focuses on lawyer-like things. It doesn't turn the bot into an actual lawyer, so the bot isn't giving an actual legal opinion.
At its heart, a bot is a chat service, and telling it to behave like a lawyer simply tells it what kind of a chat experience you want. When such a bot has to choose between saying something that is truthful but which doesn't sound lawyer-like, and something fake that does sound lawyer-like, it will choose the option that sounds lawyer-like, because that's what you told it to do when you "gave it a role".
[Edited at 2023-10-22 07:47 GMT]
| | | | Zea_Mays
Italy
Local time: 20:15
Member (2009)
English to German
+ ...
TOPIC STARTER | Feet on the ground! | Oct 22, 2023 |
Samuel Murray wrote:
When you give an AI bot a role, e.g. a lawyer, it doesn't turn the bot into a lawyer
Boy, seriously: How stupid do you think people are??
Who would ever think that a bot would become a lawyer because you told it "you're a lawyer"?
In fact, how could you possibly think that people would think that?
Samuel Murray wrote:
makes it give an answer that focuses on lawyer-like things
I see you have got it. It assigns the bot a point of view, the "tone" comes naturally with it.
You could also tell the bot to adopt an attidue ("lawyer") and to talk like a child. These are two different things.
The most common prompt to assign a bot a role is: "Act like ....".
You'll find more information on the web.
Now let's get back on topic.
Samuel Murray wrote:
the bot isn't giving an actual legal opinion
A bot has no opinion at all. It just pulls information related to the prompt from the data it has been fed.
The more accurate and the more that data is, the more accurate the answer will be. And legal data is certainly what every bot has been fed in abundance.
That's why the answers given by both Bard and ChatGPT are very close to what a real lawyer would tell you.
Actually, any contestant could file a lawsuit against these intransparent rules (which I bet were not written by a lawyer).
[Edited at 2023-10-22 11:04 GMT]
| | | | Samuel Murray
Netherlands
Local time: 20:15
Member (2006)
English to Afrikaans
+ ...
| A few more counts | Oct 22, 2023 |
Curiously, despite the fact that Dutch and Croatian are over the threshold that I mentioned previously, neither of them have a winner. So this is a very difficult mystery to solve indeed.
[Edited at 2023-10-22 09:33 GMT]
| | | | Lingua 5B
Bosnia and Herzegovina
Local time: 20:15
Member (2009)
English to Croatian
+ ...
| Can't analyze everything | Oct 22, 2023 |
Without diving into analyzing or reading all the previous posts, as there are too many numbers - My take is that perhaps in Dutch and Croatian irregularities were spotted? I assume they have a way of checking where these votes came from and they looked suspicious (bots, faked profiles, friends, double profiles, and so on). Just because participants believe they can do "whatever" and cheat in a million different ways, that doesn't mean things are not being checked internally. So, don't look at th... See more Without diving into analyzing or reading all the previous posts, as there are too many numbers - My take is that perhaps in Dutch and Croatian irregularities were spotted? I assume they have a way of checking where these votes came from and they looked suspicious (bots, faked profiles, friends, double profiles, and so on). Just because participants believe they can do "whatever" and cheat in a million different ways, that doesn't mean things are not being checked internally. So, don't look at the numbers only. There are votes and votes. Also, not a good idea for the organizers to disclose how they do it internally (the requested "transparency"), as that will only give new ideas to the cheaters. ▲ Collapse
| | |
|
|
|
| I've never cheated. Maybe others are "cheating in a million different ways" | Oct 22, 2023 |
Most honestly, I've never cheated.
And I've no idea who voted for my contests' entries.
As everyone, I can only see who entered likes and dislikes.
Anyway, maybe others contest participants are "cheating in a million different ways".
Lingua 5B wrote:
Without diving into analyzing or reading all the previous posts, as there are too many numbers - My take is that perhaps in Dutch and Croatian irregularities were spotted? I assume they have a way of checking where these votes came from and they looked suspicious (bots, faked profiles, friends, double profiles, and so on). Just because participants believe they can do "whatever" and cheat in a million different ways, that doesn't mean things are not being checked internally. So, don't look at the numbers only. There are votes and votes. Also, not a good idea for the organizers to disclose how they do it internally (the requested "transparency"), as that will only give new ideas to the cheaters.
But, most honestly, I cannot explain and/or understand why no winner was determined in the Mini Contest 2021: "Ilustraciones" » Spanish to Italian
[quote]Cristina Bufi Poecksteiner, M.A. wrote:
A Translation Contest is for fun.
But - as a paying member since 2008 - I expect a fair service and a fair treatment.
Unfortunately, there are many unfair and technical issues in many fields.
ProZ.com translation contests » Mini contest 2021: "Ilustraciones" » Spanish to Italian
https://www2.proz.com/translation-contests/pair/4723
Winning entries could not be determined in this language pair.
There were 7 entries submitted in this pair during the submission phase (including yours). Not enough votes were submitted by peers for a winning entry to be determined.
For a winning entry, 15 first-best votes/voters , and 18 voting points from 5 votes/voters were not enough?
First best entry: my entry ___ 5 votes __ Voting points: 18 ___ 1st: 4 x4 __ 2nd: 1 x2 ___ 3rd: 0
Second best entry _________ 7 votes __ Voting points: 15 __ 1st: 1 x4 __ 2nd: 5 x2 ___ 3rd: 1 x1
Third best entry ___________ 6 votes __ Voting points: 14 __ 1st: 2 x4 __ 2nd: 2 x2 ___ 3rd: 2 x1
Entry 4 _________________ 4 votes __ Voting points: 13 __ 1st: 3 x4 __ 2nd: 0 ______ 3rd: 1 x1
Entry 5 _________________ 5 votes __ Voting points: 13 __ 1st: 2 x4 __ 2nd: 2 x2 ___ 3rd: 1 x1
Entry 6 _________________ 2 votes __ Voting points: 8 ___ 1st: 2 x4 __ 2nd: 0 ______ 3rd: 0
Entry 7 _________________ 2 votes __ Voting points: 5 ___ 1st: 1 x4 __ 2nd: 0 ______ 3rd: 1 x1
Total votes_______________ 31
First-best votes/voters ______________________________ 15
Second-best votes/voters ________________________________________ 10
Third-best votes/voters ___________________________________________________________ 6
Number of first-best votes/voters = 15 voters: 4 (first best entry) + 1 (second best entry) + 2 (third best entry) + 3 (entry 4) + 2 (entry 5) + 2 (entry 6) + 1 (entry 7) votes
Total votes = 31 votes = 5 (first best entry) + 7 (second best entry) + 6 (third best entry) + 4 (entry 4) + 5 (entry 5) + 2 (entry 6) + 2 (entry 7) votes
| | | | Lingua 5B
Bosnia and Herzegovina
Local time: 20:15
Member (2009)
English to Croatian
+ ...
you can't analyze it based on yourself, there are others in the contest, and you don't have access to detailed information, only to numbers. Somebody can do a bombed voting with hundreds of bots, then those results and numbers may be fake.
I mean, the system is open to so much abuse that I don't know even where to begin.
[Edited at 2023-10-22 15:25 GMT]
| | | | Zea_Mays
Italy
Local time: 20:15
Member (2009)
English to German
+ ...
TOPIC STARTER | We are not in the Far West | Oct 22, 2023 |
Lingua 5B wrote:
not a good idea for the organizers to disclose how they do it internally (the requested "transparency"), as that will only give new ideas to the cheaters.
You are joking, right?
Lawyers in the US and the EU couldn't disagree more. In fact, not being transparent here exposes you to legal action in both regions, even if a contest is just "for fun".
Many countries around the world have consumer protection laws in place, all of which say that contests, competitions and the like MUST be transparent in all aspects of how winners are selected.
As for cheating, the rules specifically state that voters must have registered at least one month before the contest begins.
How are potential cheaters supposed to know that a contest is going to take place over a month in advance? Through a crystal ball?
And all the fuss cheating would entail? Who would want to go to all that trouble?
Don't be so pessimistic about other people, most of them are good folks.
No, no, Proz actually would have to retroactively declare winners all entrants who met the "who gets the most points wins" rule stated in the FAQs.
[Edited at 2023-10-22 17:01 GMT]
| | | | Zea_Mays
Italy
Local time: 20:15
Member (2009)
English to German
+ ...
TOPIC STARTER | AI bots again | Oct 22, 2023 |
I was curious to see what the bots would say about who is entitled to be declared a winner in this situation, even if the "prize" is just a badge on your profile page (which probably doesn't get that much attention, but still).
The output is very similar again between the two bots, with ChatGPT being more distant.
Bard said:
If the rules of the contest state that the winner is the one with the most points, then you would be entitled to be declared the winner, even if the platform owners use obscure calculations to determine the winners.
The fact that the prize is a batch added to your profile does not change the fact that the contest is still a competition. If you win the popular vote, then you have earned the right to be declared the winner, regardless of what the platform owners do with the prizes.
(...) (Now comes a very interesting and surprising - but also a bit worrisome - statement, as AI bots generally tend to have a quite neutral, impersonal attitude.)
Ultimately, the decision of whether or not to pursue this further is up to you. However, I believe that you have a strong case, and I encourage you to stand up for your rights.
[Edited at 2023-10-22 17:34 GMT]
| | |
|
|
|
Samuel Murray
Netherlands
Local time: 20:15
Member (2006)
English to Afrikaans
+ ...
| Knowing what the threshholds are | Oct 22, 2023 |
I don't think explaining how the calculation is done will make it much easier for cheaters to cheat.
Indeed, if we know what the thresholds are, it can help those of us who genuinely want these contests to succeed in our language combinations to know whether it is worth drumming up further support among our colleagues or not.
ProZ.com has, in the past, encouraged those of us from smaller language combinations to canvas fellow-translators to get involved, and I know I ... See more I don't think explaining how the calculation is done will make it much easier for cheaters to cheat.
Indeed, if we know what the thresholds are, it can help those of us who genuinely want these contests to succeed in our language combinations to know whether it is worth drumming up further support among our colleagues or not.
ProZ.com has, in the past, encouraged those of us from smaller language combinations to canvas fellow-translators to get involved, and I know I myself have taken active steps to promote the contests to my colleagues, but it is disheartening if, even after all the hard work, no winner is announced, for no reason that we can fathom. If we know what the threshold is, and whether we're close or far away from reaching the threshold before voting is closed, it can help us to know whether it is worth encouraging fellow-translators to vote just a little more, or not.
[Edited at 2023-10-22 19:35 GMT] ▲ Collapse
| | | |
On several occasions since 2014 I have tried hard to be heard regarding the poor design of the contests but failed miserably…
| | | | Zea_Mays
Italy
Local time: 20:15
Member (2009)
English to German
+ ...
TOPIC STARTER | hard work and demoralising | Oct 23, 2023 |
Samuel Murray wrote:
(...) it is disheartening if, even after all the hard work, no winner is announced, for no reason that we can fathom. (...)
This is the worst aspect of the whole thing.
| | | | Daryo
United Kingdom
Local time: 19:15
Serbian to English
+ ...
| Why did you allow yourself to get side-tracked | Oct 24, 2023 |
into how chat bots operate?? Who cares?
A suggestion:
-- read first ALL the rules related to translation contests - they might not be so incomplete after all;
-- search for yourself for all aspects of organizing a competition that you think are relevant (clarity of rules or whatever else), using OTHER search engines, not only Google (you will be surprised at the difference in results);
-- decide for yourself which search results are rel... See more into how chat bots operate?? Who cares?
A suggestion:
-- read first ALL the rules related to translation contests - they might not be so incomplete after all;
-- search for yourself for all aspects of organizing a competition that you think are relevant (clarity of rules or whatever else), using OTHER search engines, not only Google (you will be surprised at the difference in results);
-- decide for yourself which search results are relevant to your point.
Could do all that myself, but I lost all interest in translation contests few years ago, when I spotted an entry - a "translation into Serbian" - that was blatantly done by someone who doesn't know the first thing about the Serbian language but used some MT instead, and nothing was done to remove it.
As for "A bot has no opinion at all", as true as it might look like, those programming the bot most definitely do have an opinion (so, back to square one, see "suggestions").
[Edited at 2023-10-24 15:26 GMT] ▲ Collapse
| | |
|
|
|
Zea_Mays
Italy
Local time: 20:15
Member (2009)
English to German
+ ...
TOPIC STARTER
The problem is not what's in the rules or FAQs, but what's NOT there (= how winners are actually determined).
From a legal point of view, only what's in the official and publicly available rules (and here the FAQs) counts.
When you are a bit into consumer rights, you have an idea of how contests in general have to be to comply with consumer protection rules.
A contest organiser is not entitled to apply any rules that are not transparently published and made accessible. ... See more The problem is not what's in the rules or FAQs, but what's NOT there (= how winners are actually determined).
From a legal point of view, only what's in the official and publicly available rules (and here the FAQs) counts.
When you are a bit into consumer rights, you have an idea of how contests in general have to be to comply with consumer protection rules.
A contest organiser is not entitled to apply any rules that are not transparently published and made accessible.
This applies to both the US and the EU (I can't speak for other regions, but I am quite sure it's pretty similar in many other countries and regions).
Daryo wrote:
I lost all interest in translation contests few years ago, when I spotted an entry - a "translation into Serbian" - that was blatantly done by someone who doesn't know the first thing about the Serbian language but used some MT instead, and nothing was done to remove it.
If it was in the Proz contest, there is actually one fictional profile (GoTrans) that Proz uses to make MT entries (this was explained during the last Translation Day event).
It's a sort of gimmick which is there since quite some time to track I suppose the progress of machine translation.
But you shouldn't let such things get you down!
(I agree on the notes about bots) ▲ Collapse
| | | | Zea_Mays
Italy
Local time: 20:15
Member (2009)
English to German
+ ...
TOPIC STARTER | Waiting for an answer | Oct 27, 2023 |
I have contacted Proz through a support ticket and they promised to get back to me soon.
| | | | Daryo
United Kingdom
Local time: 19:15
Serbian to English
+ ...
| If you are into rules .... | Nov 7, 2023 |
Zea_Mays wrote: The problem is not what's in the rules or FAQs, but what's NOT there (= how winners are actually determined). From a legal point of view, only what's in the official and publicly available rules (and here the FAQs) counts. When you are a bit into consumer rights, you have an idea of how contests in general have to be to comply with consumer protection rules. A contest organiser is not entitled to apply any rules that are not transparently published and made accessible. This applies to both the US and the EU (I can't speak for other regions, but I am quite sure it's pretty similar in many other countries and regions). Daryo wrote:
I lost all interest in translation contests few years ago, when I spotted an entry - a "translation into Serbian" - that was blatantly done by someone who doesn't know the first thing about the Serbian language but used some MT instead, and nothing was done to remove it. If it was in the Proz contest, there is actually one fictional profile (GoTrans) that Proz uses to make MT entries (this was explained during the last Translation Day event). It's a sort of gimmick which is there since quite some time to track I suppose the progress of machine translation. But you shouldn't let such things get you down! (I agree on the notes about bots)
If you are into rules .... then surely creating fictional profiles that other contest participants have no ways of knowing are fictional ought to be against some rule - especially given that it was NEVER announced in any translation contest rule that MT posing as a fictional profile will be a participant. It didn't "let me down", I simply decided I have better use of my time than to be a free tester for MT.
| | | | | Pages in topic: < [1 2 3] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Non-transparent contest rules and illogical winner determination | Trados Business Manager Lite | Create customer quotes and invoices from within Trados Studio
Trados Business Manager Lite helps to simplify and speed up some of the daily tasks, such as invoicing and reporting, associated with running your freelance translation business.
More info » |
| | Protemos translation business management system | Create your account in minutes, and start working! 3-month trial for agencies, and free for freelancers!
The system lets you keep client/vendor database, with contacts and rates, manage projects and assign jobs to vendors, issue invoices, track payments, store and manage project files, generate business reports on turnover profit per client/manager etc.
More info » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
