cooperating voting systems in parallel

[moolight-seashell]

Add the possibility to have several voting systems working together.

For example for youtube, we could have a voting system by video (as currently) but also by channel, with a vote comparing the different channels. But also with a voting system for music videos which are completely rejected by the current system

and maybe a system allowing to rate individually the videos or channels possibly for the most positively or negatively rated videos to define a scale

I find that we have a lot of good video with a negative mark on the website, in opposite the first video of this month the most useful , is a video of 1h30 on tournesol , hmmm , It doesn't look very representative

I can be wrong but if I understand well, if on the platform people propose only good videos, by the game of comparison some will be noted negatively whereas they are all good videos, so does we need the same quantity of good and bas video to have a representative score ? and what is the result of this if all video are bad we will have video with good mark ?

so maybe add an option to add automatically all wached video in the compar list ?

[GresilleSiffle]

Hello @jean-raphael

Thank you for taking the time to give us feedback :+1:

As a Tournesol developer I'm going to give you my personal opinion. I'm speaking for myself so don't take it as the official Tournesol team's point of view.

I'll try to be factual when I can.

But also with a voting system for music videos which are completely rejected by the current system

and maybe a system allowing to rate individually the videos or channels possibly for the most positively or negatively rated videos to define a scale

I find that we have a lot of good video with a negative mark on the website, in opposite the first video of this month the most useful , is a video of 1h30 on tournesol , hmmm , It doesn't look very representative

The first question I ask myself when I read this is: why? If the motivation is not clear, the purpose of such a system in term of public interest won't be clear too.

Currently we don't plan to add new compare-able entities (like channels) in short or medium term. The first reason is because we are focused on improving the user experience on the existing videos poll, but comparing channels is something we are thinking about.

The second question is what is a recommended channel? and what is a non-recommended channel? Is it a channel with a lot of well rated videos on Tournesol? And what is "well rated"? high score? high average score? a high median score?

If a recommended channel is a channel with well rated videos, we can already display such a list in Tournesol. We don't need to ask the users to compare channels if they have already compared their videos. If a recommended channel is not deductible from the current data we have, we might need to create a dedicated poll.

Creating a dedicated poll might be a delicate operation as the results of the channels poll may not be coherent with the results of the videos poll.

I personally don't like that much the idea of comparing channels, or even creating a ranking based on the Tournesol data. Here is my humble opinion:

a ~ I think it will hurt a lot the diversity of the recommended information. Currently the top videos of Tournesol come from different channels and are supposedly recommended regardless of their author. This is because Tournesol focus on the information rather than on its origin.

b ~ I think it can shadow a lot information of public interest. Some YouTube channels are focused on a specific topic, and not all videos are video of public interest, but some of them are. What if a YouTube channel has hundred if non-recommended videos on Tournesol, but has the one that is the most recommended on Tournesol? Depending on how we compute the score of this channel, we may shadow this great video.

c ~ Videos are information, channels are groups of videos of different quality. Recommending videos to people may create a better world, a world where people will take better decisions. Recommending channels to people is in my opinion far less reliable. Are we expecting the users to watch all videos of those recommended channels?

d ~ Comparing videos is sometimes difficult, comparing channels might be even more.

To sum up, I'm more in favor in displaying a "funny" horizontal ordered list of channels, on several lines, based on the current Tournesol data (just for pleasure of viewing stats), rather than displaying a vertical ranking of channels in a dedicated poll inviting people to check the top channels (because it may hurt the diversity of information).

and maybe a system allowing to rate individually the videos or channels possibly for the most positively or negatively rated videos to define a scale

What do you mean exactly by rejected? For instance I'm able to add musics on Tournesol, and these videos are not rejected.

If they fall into the non-recommended side of Tournesol, it doesn't mean these are bad videos, it means they have lost theirs duels against more recommendable videos.

This is a very important feature of Tournesol we might need to explain somewhere (in the Wiki for sure, but maybe somewhere in the application).

If I remember correctly we plan to have an extra feature allowing users to mark a video as recommendable, so that it's minimum score is 1. But I don't know if it's still on the road map.

I can be wrong but if I understand well, if on the platform people propose only good videos, by the game of comparison some will be noted negatively whereas they are all good videos, so does we need the same quantity of good and bas video to have a representative score ? and what is the result of this if all video are bad we will have video with good mark ?

Users are indeed invited to give their opinions on which video should be the most largely recommended, this is currently our definition of "good video".

Losing a comparison for a video doesn't mean it will have a negative score. For instance if a video A has a great score, and video B is rated a bit less recommendable, B will be considered very recommendable and will have a positive score.

and what is the result of this if all video are bad we will have video with good mark ?

Exactly. The top videos will be the most recommendable videos of the "bad" videos. Hopefully YouTube contains a lot of good content : ) That's my assumption.

so maybe add an option to add automatically all wached video in the compar list ?

Yes we definitively need to work on how users can add videos to theirs rate-later lists. Thanks for the suggestion, we will think about it.

[moolight-seashell]

ok thanks

by rejected i mean, because music is never fundamentally important, the voting system encourages people to vote the against any music video. That's why I imagined a parallel sunflower with a music rating

this is currently our definition of "good video".

ok i understand the concept but it's far from my intuitive definition of a good video,

In fact, if I understand correctly, this system emits statistically as many positive and negative notes which is quite arbitrary and not very representative of the qualities and defects of the videos.

And this average of the notes to zero seems to me a little far from the idea that I have of a good video

but ok i get it