Boards›Player Behavior

The instant IFS system.

Fresh Untilted·10/17/2017, 4:13:27 PM·3 votes·665 views

The idea is to make the IFS, the Instant Feedback System, well...instant. After all, you can't teach a dog to behave unless you scold them right away. The same thing applies to humans, to an extent.

Whenever you say a blacklisted phrase ingame, your chat will not be sent, and you are warned by a yellow message that only you can see (appears in chat). The phrase can be something like "I wouldn't say that if I were you ;)" or something else that deal with their negativity in a semi-positive way.

The way it would work is this:

Player A reports player B. Riot's systems review the game, and find several trigger words or phrases that appear frequently in the chat logs of those who are reported for abusive chat. If this behavior is enough for a suspension or restriction, the phrase that triggered this punishment will be blacklisted. This can include as much as a sentence, or as little as a word. This way we know:

  1. Player A was offended by it, enough to report.
  2. Riot was offended by it, enough to punish.

This equals to the banning of only those phrases that directly influences whether someone is punished. In this way, normal phrases that Riot never deems toxic enough to punish, are left alone.

After a year of this system, the system will cross-reference those phrases with hundreds of thousands of reports. It will be really hard to keep inventing new ways to be inflammatory. Those who still find a way, are punished accordingly (also easier to punish people who try to circumvent the blacklist by adding numbers or letters to phrases), and the phrases they used are prevented from being said in future games.

It will also make it easier for Riot to review chat logs. The system could highlight the phrases that appear in a fair number of reports. Riot will go through them, deem them toxic enough for punishment, and then mark them for blacklisting. Apart from reports, the only manual part of this system would be made quite a lot easier, giving the IFS team to work on keeping the blacklist relevant.

Lets take a few examples:

Player A1-1000 says in their chat: "I don't like idiots". The entire phrase is banned.

Player B says: "I don't think any of you are idiots".

Player B's chat would NOT be blocked unless Riot blocks the word "idiots". He would for instance be allowed to say "idiots" in a context that does not appear in hundreds of other reports. However, if he is punished for saying for instance:

"Idiots like you should uninstall", and that is a new phrase that still gets through to Riot by being in the background of a chat log (unhighlighted), the phrase would still be banned if marked by Riot.

That means, the longer Riot has this system, the better it becomes at preventing toxic chat. Sorry for rambling on for so long. I appreciate you took the time to read it.

20 Comments

Xidphel10/17/2017, 5:04:12 PM5 votes

The instant IFS system

INSTANT INSTANT FEEDBACK SYSTEM SYSTEM

ModThe Djinn10/17/2017, 4:17:24 PM2 votes

I think this is a little too complicated to realistically do, and fairly easy for people to work around once the system's limitations are known.

Additionally, it misses a whole bunch of toxicity that's created by repeated minor harassment, or by using important game information to belittle other people.

For example: it's fine for me to say "Don't fight him. He has four kills on you, and will almost certainly win."

It's not okay for me to keep saying "four kills" "five kills" "six kills" every time my teammate dies to rub their failure in their face.

Kei14310/17/2017, 5:08:39 PM2 votes

a neat idea, probably hell to program.

The Sword Saint10/17/2017, 8:37:10 PM1 votes

A neat concept, but it seems like it may be overengineered. To create something like this that actually works the way you envision would likely be the kind of thing that universities write thesis about and request for research funds in order to aim for a headline grabbing nobel prize.

A simpler approach is probably available to hit most of your main concerns - something like a new blacklist that is periodically expanded by hand based on machine gathered data combined with a chat filter that prevents sending messages in game that match some percentage against the blacklist. There's already software that does this to compare lines of text against databases of published work and websites to check for plagiarism, so the tech is there to do this. It wouldn't be as unreasonable for riot to implement such a feature and update the blacklist a few times a season to keep up with hip new trends.

Telephone Booth10/18/2017, 3:20:18 AM1 votes

Lol the instant instant feedback system system.

On the subject; I am far too ignorant to understand coding and machine learning, so I can't really have an opinion on this.

Drugoth10/18/2017, 2:16:18 PM1 votes

In my own opinion, I don't think the IFS censoring certain phrases automatically is strictly needed. It feels like if anything, this proposed change would benefit flamers in that it could prevent them from flaming, therefore protecting them from themselves.

I think more responsibility needs to be put on the player to govern their own actions, rather than just protecting them from themselves.

It's like bowling with the bumpers up. Sure, you can't get any gutter balls, but it also doesn't encourage you to learn to keep the ball in the center of the lane. You see what I mean?