The Speech to Text feature tells us that Rare is already transcribing audio, and already can identify bad language (as it is ***'d out), and all audio goes through their servers already, so it can also be recorded tied to the STT text. Certain language is always against TOS, (e.g. n-word). It feels like Rare could set up a system to flag anti-TOS language and trap the associated recording, and then have one person on their end sitting there reviewing each audio clip manually and clicking ban. With this sort of setup, they could easily ban at a rate of a few per minute. This would be such an easier way to handle audio TOS violations and clean up the seas than having players go through all the work to record and report it themselves.
Streamline banning using Speech to Text and recorded Voice Comms
The Speech to Text feature also tells us that their accuracy can be pretty low, especially with certain accents which makes trying to automate banning sketchy at best and would likely create a lot of false positives. Simply saying a bad word isn't a bannable offense; context matters.
There is no way they are going to have people listening 24/7 to millions of audio recordings and manually banning.
@d3adst1ck Sure Speech to Text isn't 100% accurate, but it is going to catch a bunch of the bad stuff, and I specifically suggested a manual review of the STT triggers, not automated bans or manual review of all voice comms, it's the intersection of these two that I am proposing. I am not sure how many of these will trigger every hour or every day, but they could set it up such that those with the most repeat STT violations are at the top of the manual review list. If they put a bunch of effort into reviewing/banning in the first few weeks, people would get the message quickly and the total violations would drop pretty quickly. It just seems like a much more efficient way of directing their review/ban efforts, and reducing the amount of toxic stuff that players have to experience and then report before it gets dealt with.
@wolfmanbush Sure, but if you were automatically flagged for a TOS audio, the manual review will show if you are really a TOS violation or not, so there is nothing to worry about, and there is still plenty of easily recognizable audible TOS violations that I see regularly either playing myself or watching other streamers. It feels like these should be able to be handled by Rare without the need for other-player reporting.
@slarow said in Streamline banning using Speech to Text and recorded Voice Comms:
@wolfmanbush Sure, but there is still plenty of easily recognizable audible TOS violations that I see regularly either playing myself or watching other streamers. It feels like these should be able to be handled by Rare without the need for other-player reporting.
Imo sea of thieves has minimal toxicity when it comes to serious violations of any reasonable ToS. It happens, people will see it but imo it's less than many similar shared environments.
What I think SoT does have a lot of is a pride and hypocrisy in tattle talin'. A lot of baiting and hunting for problems.
I'm not talking reporting serious violations I'm talking contributing to an environment where reporting and escalating are a weapon for perceived social reward, Like chasing down an upvote but in a way that leads to more negativity.
If people have an issue they can go through the process of reporting with evidence. Trying to monitor everyone for complete control isn't a way to hold together a healthy shared environment imo.
10k+ hours I've seen more people trying to use it as a weapon for perceived personal/social gain than I have a tool to remove people that are engaged in clear behavioral toxicity.
This suggestion is going to create more issues than it solves, just like escalation on the seas does.
Getting many people caught up in a position of needing to defend themselves over flawed and inaccurate technology doesn't help much of anything.
The people that truly want to hurt others with words and are competent in causing harm are going to use condescending passive aggressive attacks anyway and that isn't going to trigger filters. The wild over the top and less strategic ones get caught pretty quick already.
@slarow said in Streamline banning using Speech to Text and recorded Voice Comms:
@d3adst1ck Sure Speech to Text isn't 100% accurate, but it is going to catch a bunch of the bad stuff, and I specifically suggested a manual review of the STT triggers, not automated bans or manual review of all voice comms, it's the intersection of these two that I am proposing. I am not sure how many of these will trigger every hour or every day, but they could set it up such that those with the most repeat STT violations are at the top of the manual review list. If they put a bunch of effort into reviewing/banning in the first few weeks, people would get the message quickly and the total violations would drop pretty quickly. It just seems like a much more efficient way of directing their review/ban efforts, and reducing the amount of toxic stuff that players have to experience and then report before it gets dealt with.
What you consider 'bad stuff' is just words. Two friends could be playing on the same crew and communicate using colorful language. Or it could not be and the audio is just poor and thinks they said something bad, as evidenced by the number of times regular words get ****'d on text chat. The comments are not malicious, and both players know that because that is how they communicate but the person reviewing the audio doesn't - they just know that a bad word was said and do not have the context behind these players, their history or the current session. Player who said the bad word gets banned because you've implemented an automated system the removes contextual nuance and relies only on a strict set of blacklisted words to determine what is and isn't an offense.
Ignoring the insane amount of manpower you'd need to add to manually review everything that is triggered and attempting sort out what is actually actionable using what little context remains, this is a poor system and is why no company does this and the ones that do ban based on text chat triggers usually end up with a number of false positives and frustrated players.
If someone is breaking ToS, report them. If you don't care or it doesn't bother you, don't report them. This results in a much higher percentage of actually actionable reports and less noise to sort through.
@wolfmanbush said in Streamline banning using Speech to Text and recorded Voice Comms:
What I think SoT does have a lot of is a pride and hypocrisy in tattle talin'. A lot of baiting and hunting for problems.
I'm not talking reporting serious violations I'm talking contributing to an environment where reporting and escalating are a weapon for perceived social reward, Like chasing down an upvote but in a way that leads to more negativity.
10k+ hours I've seen more people trying to use it as a weapon for perceived personal/social gain than I have a tool to remove people that are engaged in clear behavioral toxicity.
So wouldn't it make sense to put more effort into directly addressing the issue from Rare than relying on users to report toxicity? Automated triggering and recording of the offense and the surrounding context with manual review has much less opportunity for abuse than users manually recording toxic behaviour.
@d3adst1ck said in Streamline banning using Speech to Text and recorded Voice Comms:
What you consider 'bad stuff' is just words. Two friends could be playing on the same crew and communicate using colorful language. Or it could not be and the audio is just poor and thinks they said something bad, as evidenced by the number of times regular words get ****'d on text chat. The comments are not malicious, and both players know that because that is how they communicate but the person reviewing the audio doesn't - they just know that a bad word was said and do not have the context behind these players, their history or the current session.
If the language is used via in game chat, it doesn't matter if two friends are communicating using colorful TOS-violating language, that language could be heard by other crews and is a TOS violation.
@d3adst1ck said in Streamline banning using Speech to Text and recorded Voice Comms:
Player who said the bad word gets banned because you've implemented an automated system the removes contextual nuance and relies only on a strict set of blacklisted words to determine what is and isn't an offense.
Which isn't what I suggested, I suggested a manual review of accounts that are repeatedly flagged by the automated STT filters.
@slarow said in Streamline banning using Speech to Text and recorded Voice Comms:
@wolfmanbush said in Streamline banning using Speech to Text and recorded Voice Comms:
What I think SoT does have a lot of is a pride and hypocrisy in tattle talin'. A lot of baiting and hunting for problems.
I'm not talking reporting serious violations I'm talking contributing to an environment where reporting and escalating are a weapon for perceived social reward, Like chasing down an upvote but in a way that leads to more negativity.
10k+ hours I've seen more people trying to use it as a weapon for perceived personal/social gain than I have a tool to remove people that are engaged in clear behavioral toxicity.
So wouldn't it make sense to put more effort into directly addressing the issue from Rare than relying on users to report toxicity? Automated triggering and recording of the offense and the surrounding context with manual review has much less opportunity for abuse than users manually recording toxic behaviour.
This to me is like saying shouldn't we just record all public areas in the neighborhood and have people with power watch over everything to weed out people that take pride in reporting.
Replacing a flawed but necessary system with a highly intrusive system isn't improvement it's just a more intrusive shift of power and flaws.
I personally value privacy significantly as often as possible including (to some degree) in public areas. Privacy takes a hit during an experience like SoT but just because it takes a hit it doesn't justify full surveillance mode of individuals.
For me personally I'm not even fond of what goes on with streaming. It's a public space but imo taking someone that may be saying something that would embarrass them or cause some sort of amplified stress by blasting it to (potentially) thousands of people is not a responsible way to create content. It's common in everything now but I don't think that justifies it. Personally I think there should be some sort of option to where people can opt out of having their gamer id broadcasted to the world (only the streamer currently decides), even in a online environment. They can still create a unique session id (tied to the account) that can be used for reports.
I don't think people need to be monitored constantly.
@slarow said in Streamline banning using Speech to Text and recorded Voice Comms:
If the language is used via in game chat, it doesn't matter if two friends are communicating using colorful TOS-violating language, that language could be heard by other crews and is a TOS violation.
No, it's not. They even specifically mention this in the Code of Conduct:
Be courteous. Don’t insult players you don’t know. If humour is the intent, remember that this can be lost or misinterpreted online, and words can very easily offend. Be mindful of what you’re saying and if someone asks a question, see if you can help them out as you’d appreciate being helped out yourself if the roles were reversed.
Keep it appropriate. Avoid swearing in unfamiliar company where possible. Don’t post anything containing NSFW material – this has no place in Sea of Thieves or its community spaces. And while we all have opinions of the world around us, people aren’t necessarily playing Sea of Thieves to talk about politics or contentious real-world issues. Switch off, escape and enjoy the freedom of the seas instead.
They would have said do not swear, period if any use of bad words regardless of context would get you banned. Context matters.
@d3adst1ck said in Streamline banning using Speech to Text and recorded Voice Comms:
They would have said do not swear, period if any use of bad words regardless of context would get you banned. Context matters.
I am not talking about general swearing, I am talking about the next level of words/comments that are going to be TOS violations whenever used.
There's already technology in place in some games (not sure if SOT uses it or plans to) that analyses voice comms in games for potentially toxic encounters, it then directs enforcers to have a closer look (listen in).
For those wondering about privacy issues, I think you'll find somewhere in the Xbox TOS that you waived any concerns with that.
I have the speech to text on in SOT anyway so I can tell who is talking at any point, should I need to report incidents (slurs or extreme toxicity), which thankfully isn't often. It's helpful for identification of who is speaking but the accuracy of the text translation is often questionable. My accent/voice is especially hard for the speech to text to pick up correctly.
@slarow said in Streamline banning using Speech to Text and recorded Voice Comms:
@d3adst1ck said in Streamline banning using Speech to Text and recorded Voice Comms:
They would have said do not swear, period if any use of bad words regardless of context would get you banned. Context matters.
I am not talking about general swearing, I am talking about the next level of words/comments that are going to be TOS violations whenever used.
In order for your automated system to work, it has to pick up general swearing because you can use those words in a way that violates ToS if directed at another player. This means that even conversations between two players with no one else around would become actionable, even if the way those words were used was not in any way actionable and was previously acceptable under the Code of Conduct.
While I know that this idea is rooted with good intentions, it feels a bit too "big brother" to me.
A few of my friends are what we would consider "potty mouths" and would likely be constantly flagged...even though the foul language is only amongst friends.
I get it...probably shouldn't say certain things regardless of whether or not it's amongst friends...but that's just not how the world works.
Above it all...the biggest challenge is that Rare would need a 10+ person multi-lingual team reviewing what I would imagine would be a LOT of audio clips.
I get it...it's got the best of intentions...but, you're talking about adding a huge resource to a relatively small gaming company like Rare.
@d3adst1ck said in Streamline banning using Speech to Text and recorded Voice Comms:
The Speech to Text feature also tells us that their accuracy can be pretty low, especially with certain accents which makes trying to automate banning sketchy at best and would likely create a lot of false positives. Simply saying a bad word isn't a bannable offense; context matters.
There is no way they are going to have people listening 24/7 to millions of audio recordings and manually banning.
The solution to this is only making certain words (like the n word, f slur, you know what I mean) trigger the automatic report.
Maybe not auto ban but for example xbox has a built in option prompting a report so easier method of reporting abusive text, you shouldn't have to go to so much effort to provide evidence and raise tickets external to the game to make the community a better place when like you say there's logs, esp when some of it is typed so def intentional and not translated speech to text
@vprogamer1086 said in Streamline banning using Speech to Text and recorded Voice Comms:
Also “cleaning up the seas” will never happen until all cussing is taken out of game, because cussing is always been nastier than “slurs” due to it being harder to apply slurs to people due to not knowing their race, appearence, or mental state.
@wolfmanbush said in Streamline banning using Speech to Text and recorded Voice Comms:
@vprogamer1086 said in Streamline banning using Speech to Text and recorded Voice Comms:
Also “cleaning up the seas” will never happen until all cussing is taken out of game, because cussing is always been nastier than “slurs” due to it being harder to apply slurs to people due to not knowing their race, appearence, or mental state.
Glad I'm not the only one who did a digital doubletake
@sweetsandman said in Streamline banning using Speech to Text and recorded Voice Comms:
While I know that this idea is rooted with good intentions, it feels a bit too "big brother" to me.
A few of my friends are what we would consider "potty mouths" and would likely be constantly flagged...even though the foul language is only amongst friends.
I get it...probably shouldn't say certain things regardless of whether or not it's amongst friends...but that's just not how the world works.
Above it all...the biggest challenge is that Rare would need a 10+ person multi-lingual team reviewing what I would imagine would be a LOT of audio clips.
I get it...it's got the best of intentions...but, you're talking about adding a huge resource to a relatively small gaming company like Rare.
Yeah that makes sense. Ok so how about a compromise. Instead, Rare uses Speech to Text to track statistics on how many times the community as a whole says "Shut up Larinna", I would love to see those stats :D.
@slarow said in Streamline banning using Speech to Text and recorded Voice Comms:
Yeah that makes sense. Ok so how about a compromise. Instead, Rare uses Speech to Text to track statistics on how many times the community as a whole says "Shut up Larinna", I would love to see those stats :D.
Oh lord that would be a large number 🤣🤣
1 out of 21
