The Racing Rules of Sailing
Limitations on Coaching or Outside Assistance in Filing a Hearing Request
100% open question here. Are there any limitations of any kind in receiving help, coaching, assistance from any source .. in filing-out and filing any hearing request for any type of hearing?
Created: 26-May-12 20:27
Comments
Format:
You must be signed in to add a comment.
The pros teach how to file a protest.
So I would say no restrictions.
PS: Feel free to categorize and discriminate between them if you want.
But what if I'm not judging at an event but a buddy calls me and asks me to help him file the hearing request to make sure the written filing meets validity?
We should also check Code of Ethics/Conduct to make sure there isn't anything listed there.
Just happy to get the protests to get rules compliance in fleets.
About help given by judges or other race officials.
There's a considerable difference, which the rules commentary books about protesting don't always make clear, between writing the minimal requirements of RRS 60.3(a), and preparing for a protest.
I appreciate Paul's reservations, but given the very minimal requirements for protest contents in RRS 60.3(a), I would suggest that any help in meeting these requirements would be all but trivial. Desirably any advice should be given by a duty judge, not by a judge that will hear a protest, but particularly at club level, as Michael has indicated I don't think there's much harm in it.
Were you canvassing whether or not 'help' provided by AI in writing a protest was permissible?
In which case, my answer is absolutely yes.
Were you concerned about AI hallucination protests?
In which case, how is AI error any different from errors competitors make witout any AI assistance?
Are you contemplating putting an AI 'wizard' into the RRoS protest form and wondering if this is OK?
EDIT: OK, saw your AI Wizard in https://www.racingrulesofsailing.org/posts/5417-limitations-on-coaching-or-outside-assistance-in-filing-a-hearing-request#comment_21349
Paul started a priviate dialog with me about his new AI filing assist and within that dialog I saw the opportunity to discuss within the forum the more broader topic of both receiving coaching/assistance (from the competitor's POV) .. and giving it (from a certified Race Official POV). I tried to keep it as open as possible so that the experience of the forum might uncover aspects that we hadn't considered.
US42 came to mind as we discussed it. The first sentence of its abstract states ...
So as you can see, US42 sets out the idea that it's fine for competitors, who have both common and conflicting interests, to help each other prior to filing a hearing request and decide to act together or individually.
Yes AI is coming. If AI helps someone get their form filled out correctly, I am ok with it. The competitor is still responsible for the content. Submitting a form with incorrect facts (such as AI hallucinations) can get the competitor into trouble. To be honest, in my experience, the content of the form aside from validity does not have a lot of impact on the hearing.
As far as helping with filing of protest is concerned by the judges, it would be fine if the judge is helping with getting the protest filed online, as in, if the competitor is unable to access the system online and you as a judge are helping with a laptop and internet. However, as a judge we should not be guiding them for the contents of the protest; which should be done by the sailor.
I'd like to press you on your position here to be more detailed and backup that position by something in the rules.
Above, is that any "judge" or are you speaking of a judge that will sit on the panel?
As John A points out, what if it's not coaching for specific content elements but only content existence? For instance, "You left the Protestee blank". .. or "Hey, you need to describe the incident specifically enough so that the PC and the Protestee identify the incident you are protesting about".
What specific element of the Rules or Codes are you basing your position on?
Above, is that any "judge" or are you speaking of a judge that will sit on the panel?
I am just referring to help to the sailor as far as providing him access to Internet or laptop in case the sailor is unable to file the protest due to connectivity or other issues. I am in no way saying that the judge needs to fill in the protest for the sailor.
On what basis in the Rules or Codes do you support your position? Why can't a judge go further? .. and are you limiting that to any person who is a qualified judge or only judges that are sitting on the panel?
I'd equate helping someone log-into a site to offering them a pencil and a piece of paper or protest form. I'd agree that any judge, even one sitting on the panel, can provide this level of assistance ... but that is an extremely low threshold.
That is to say, assist competitors to meet the minimum requirements for their protest to be valid
I guess that that is where the boundary between permissible help or facilitation, and improper coaching falls
Once ever a judge starts talking about the description of incident or diagram, they have crossed the line.
That's way out of court.
Once a protest validly gets into a hearing, judges can and should 'help' parties present their cases by asking good questions to get out all the relevant facts.
Minimum requirements to get a valid protest, no further.
OK to lend a computer if your event requires online submission. Or a pen and form if not.
Regurgitate the rules: "Make sure you say who the protestor, the protestee and what the incident was. Fill out the bits about the flag and hailing etc... You can put a diagram if you want."
Then leave them to it. Don't get involved with helping wording the description or drawing the diagram. That is tampering.
So the line for me is drawn at 'limit my assistance to what is available in RRS'.
----------------------------
The AI trail is interesting.. need more time to think about it. It is leading in some areas as it currently stands.
Impressive!
What do you think?
In addition to the ones already mentioned, some other problems with too much assistance:
There are two kinds of assistance.
Helping a party complete a form so that the rules can be enforced should always be ok. This includes providing a protestor with the tools required by the SI (paper/PC) and telling them what is required under the rules. Putting RRS 60.3(a) in even more 'layman' terms (if that is possible). Answering the question, " Which rule is the one about tacking?" is fine too.
But suggesting a protestor change their testimony to make something 'more believable' oversteps the line'.
So to me, an AI tool which checks for all the elements required to for validity is ok. Input boxes with example text in perhaps are on the verge. The full coaching AI package, describing which rules to quote and what is missing from the testimony may be a little too far, and result in skewed testimonies.
Thoughts?
Paul was concerned about:
As for helping a sailor meet the [minimum] requirements for a valid protest, that's help that is [presumably] boign to be provided to any protesting boat equally. The situation is asymmetrical: there is no corresponding help that an be provided to the protestee at this stage.
It's up to the judge to not let the conversation get that far.
When I introduce myself at a Competitors' Briefing I sometimes say
"Please feel free to come up and talk to me about rules, that is unless it is 20 minutes before you are going to be talking to me about those rules across the table in the protest room."
Whether we realize it or not, the competitors are using AI to prepare for hearings. We need to adapt, and change is always hard. One of the reasons I want to package the AI in the application is so that we can teach it through our experience with it. I'm thinking that included in the process is a place where judges can provide feedback to the AI, where they can indicate what worked or didn't, or what was inappropriate. We can then include those rules in the prompt and the AI will get smarter as people use it.
I'm completely open to hearing criticism and concerns. However, no matter how we see AI participating in the hearing process, we're already underestimating our experience with competitors who are using it. My goal here is to provide tools to everyone involved to perhaps make AI more functional and reduce some potential abuse. Let me know any thoughts you may have, but I really do want to drill down into the specific issues.
(I did say I needed to think it through and deliberately used the word 'may'....!)
I am happy to admit that also after thinking this through, I am tending to agree with your logic.
When I first saw this thread, I was immediately reminded of a case a few years ago where we received a protest from an Optimist competitor who was still out on the water racing! A well written protest description submissible in any court. (It was well known that the sailor's father was a high level courtroom barrister, and happened to be a spectator that day.) We were all surprised by this and many thought this was a step too far. That was the first of many realisations that the old way was gone in a big way.
Years later, and we are still learning the new way. This is I guess is an extension of that but with some twists.
Human vs Robot.
When discussing a human coach (member of the PC), in the past we have generally steered clear from too much intervention being appropriate. There is a sense of possible bias (or optics of bias). I guess, human nature can be a problem. Jealousy, prejudgement, natural bias and all the other emotions which make a human can be at play.
With AI there is none of that. So that problem is in effect, annulled.
I think the discussion so far (based on your OP) has intertwined a load of concepts with the fundamental enquiry about AI coaching. So I'll stick to the AI coaching discussion going forward, rather than debating whether a jury member giving advice at the jury desk is ok.
Is there a difference between rules training (abstractly done at a rules seminar by Dave Perry) and coaching to win a specific case? Perhaps, but so hard to define that it is not worth delving into.
This leaves the question of whether AI coaching 'skews' testimony.
Take the question for example, "When did you fly the flag?" A classic one. The protestor remembers they flew it about 20-30 seconds after the incident. They load up the form and find the input box pre-filled with an example word...'Immediately...' They remember something about having to fly the flag very soon after, so they simply go with the example word, rather than their memory. The form has skewed their testimony.
Can 'coaching' by AI or human skew testimony. Yes. The ability to lead a witness is a well studied topic. The question is, is it any different to what currently happens? Lawyers advise clients. Racing rules advisors and coaches advise sailors. Lying is forbidden and that is a mantra we have to trust.
So, is an AI coach any different to a human 'rules advisor' in that respect.
In short Paul, I don't see any argument against AI tools. As with many things, AI is pushing us to pardigm shifts on how we operate. It's inevitable.
So .. are we going to see 2 buttons?
1) AI assist: file a hearing request (Protest, R4R, Reopen)
2) AI assist - I've been protested, help me prepare to defend myself.
Also .. are you suggesting that in RRoS.org event system .. judges will be able to see which boats used the AI, their prompts and iterations?
Will the judges have AI analyze if the competitor iteratively changed facts until they got a solution where they were in the right?
'You have been protested. Click this link for AI help to prepare for the hearing. '
If i see a form with insufficent details, before a hearing i point this out so they have time to remedy it.
Seems sensible i hate invalid hearings.
If you disagree then what if appendix t, you say invalid.
They then remedy before the hearing, what isthe difference?
The remedy windows are now closed in the current quad. Once the PTL is past, the delivered hearing request can't be perfected to cure ailments.
I am happy to help to get a valid protest the committee can chech.
Happy to watch them write the facts.
As for the rules, i leave it to them to select or leave, i do not discuss the rules with one party.
Why cannot these defects be perfected before the hearing if the form was lodged on time.
I see now they have changed the rule, certain defects could be remedied, now there is no flexability.
A lesson learned.
I believe a forum member actually submitted it to the Answer Service and shared the response if memory serves.
Not at all.
For a start, a witness that has been 'schooled' in what to say and how to say it is highly problematic. If a coach tells a witness that such and such must have happened, then the witness may testify that it did happen, whetehr they actually saw it at the time or not.
It's often said that an eyewitness's unmediated and contemporaneous testimony is more authentic and reliable than the testimony they may later give, particularly after having had a coaching discussion that highlighted important 'facts' necessary to support an allegation of a particular rule breach.
Judges can sometimes identify 'reconstructed' evidence: this is when a witness says 'It must have been ... ' something: This is not evidence of what the witneess saw or heard, it is something reconstructed from other perceptions. Reconstructed evidence should be given less weight than direce evidence of what was seen or heard.
Whether intentional or not coaching of a party or witness can skew their evidence. However, judges can't proceed on the assumption that because there has been a prior discussion (which is perfectly permissible within the rules) then the evidence is less truthful or accurate than otherwise.
Sailing is a coached game and judges have no business, outside of RRS 41 or substantial evidence of lying in breach of RRS 69, in interfereing with any coaching process.
If a competitor asked AI 'Tell me what to say to win my protest that NNN broke RRS 13'' that would get them a specification or a script.
I don't think this would be any different from the result of asking a human coach or any other support person the same question.
I think an ethical coach or support person would be very reserved in answering that sort of question.
I'm more than happy with the expanded prompts provided in the AI Protest Form you have exhibited in this thread.
Consistent with view I've expressed that it's OK and helpful for a judge at an event to help a competitor up to the point of making a protest valid, and probably no further, I think the RRoS platform is moving beyone it's 'charter' when it starts to provide 'coaching'.
Doing a proper job of evaluating and endorsing a particular AI LLM and a particlar interrogation script would be a pretty big undertaking.
Preserving that coaching for the parties and judges to see would create a relatively voluminous written document for judges to read. I'm not sure that that would lead to 'more concises hearings'.,
If a coach or AI encourages a witness to make up or embellish evidence, that's a bad thing. It may be impossible for a judgee to detect. I guess that there is no real reason to expect that AI is any more likely to do this than a real live coach, except that accredited coaches sign up to a Code of Ethics.
Again, I don't think there's anything much judges can do about this.
I guess so, but, as I've said above, can we be sure that we've done a good job of selecting the AI we are linking to.
What about the case where a competitot accesses some other AI, and doesn't disclose it to the protest committee.
1) A judge involved in the regatta as a judge helping (and maybe also being in the hearing)
2) A judge who is not involved in the regatta or another knowledgable coach helping
3) AI helping a competitor fill out the form and maybe even giving advice on the case
A lot of the discussion has focused on case 1. Clearly there are some limits here. We may have some disagreement about where those limits are but I don't think that is relevant to what Paul really wants to know.
Case 2 in my mind (and I think in John Allan's mind if I understand him correctly) has very few limits. At the highest levels of our sport (Olympics, America's Cup etc.) teams have rules advisors. Many of those rules advisors are trained and certified as judges. They are certainly helping their sailors with writing and filing protests (and in some cases even writing the protests themselves). For sailors who do not have rules advisors to call up a friend who happens to be a judge (or not) to get advice is I believe quite reasonable. The only people who have to careful here are the judges who may be becoming a coach and therefore have a future conflict of interest. Yes the father may have written up the protest, but the Opti sailor had to represent the story in the room (assuming it involved a rule of Part 2, 3 4 on the water). All the polish in the form doesn't change that.
Case 3 I believe that this is similar to Case 2. Yes competitors will use it outside of the platform. I see no reason not to build it into the platform (assuming that it doesn't get too expensive). I don't care if the AI helps the sailors with the forms or explains the relevant rules to them. The sailor still has to represent the case in the protest room. This is the case that Paul is asking about. The system needs all of the gaurdrails that any AI system needs to avoid misuse but I see no reason to put specific limit on how much it can help the sailor.
If we ever get to the point (and we will) that AI is deciding on some protests then maybe we need to make sure the two systems are isolated from each other. This would be like Case 1 but we aren't there yet.
Like several others in the discussion I am an IJ and I work in IT so I am seeing every day the impact of AI. It is here, the sailors will use it. I see no reason (other than cost) not to build it into the system.
Without good training what results will you get
I do not believe i have received any ai protesrs.
The poblem isn't to much text it is that electronic protests are generally so sparce.
You cannot exen pre scribe from them as they have few facts.
We have to consider the form to clear off the issuees and look for inconsistency. Extra added by ai and not substantiated would cause us to doubt the evidence to the detriment of the protestor.
I am all in favour of ai making sure a protest is valid, but really not more.
It will not know the facts, just those needed to win
Coaches books etc are sought and considered, and help edit or become concise.
I see ai at the push of a button leading to lots of additional text.
No i have not looked at the example but Will be doing so.
I think facilitating AI advice to competitor on how to perform in a protest hearing is a breaking-the-sound-barrier level scope creep.
Up to now RRoS has been a portal to authoritative WS and MNA documents. It doesnt provide any 'advice', or even interpretation and examples li Ulli Finkh's sites. It doesnt provide advice like a coach would on tactics or boat speed.
Suddenly the site is getting into the advice business.
It seems to me that a separate site for advice might be a better idea.
On the one hand, the charter (not that there necessarily was one) creep is a point. Will there need to be disclaimers? Could charge more for 'AI-Rules Advisor-Pro'!
On the other hand, as someone also pointed out, if it's not done here, someone else will do it. And perhaps badly. (I did wonder if this site could link to an external AI to keep to the 'charter'. But what's the difference.)
For those of you who have been around from the beginning, you may remember the concern in the early years with using a computer during events. There were some who adamantly refused to use RRoS as it wasn't the "pure" way of running a regatta and they didn't want to include the use of computers. For better or worse, change happens. I think nearly every competitor under the age of 25 is using AI routinely, whether or not you realize it. And honestly, we're a little late in implementing some AI solutions as the technology is pretty broadly integrated.
Having said that, thank you for the discussion. It has been helpful to talk and think about the issues involved. When the AI features are deployed, we'll continue to adjust and refine the process as we receive feedback. But I think this discussion has given me some ideas around how and why it will be used.
The move to directly assisting the competitors beyond fine-tuning and improving the user-friendly UI might appear at first to be a 'creep'. However, this may just be imagined.
I, for one am always impressed at this site as a management tool, and the way it has been developed over the years as needs and ideas come up. I think the incorporation of AI is no different and can only enhance everyone's experience. It's an inevitable requirement.
The AI protest form is front-of-house, that's true. Direct competitor assistance in this way is new. New technology always will need time to be 'thought through'. For me, it was a matter of hours. For others it might take longer. (Some may never accept it.)
Interestingly, I wonder if AI use will emerge most valuable behind the scenes or to complete mundane tasks. As is often the case. Many uses are not yet realised and waiting to be discovered over time.
So there's no doubt about it. AI is the future - whether used to assist competitors or officials or both. The possibilities are endless - must start somewhere.
Good job Paul.
-------------------------
An AI prompt for OTW pairing, I used at an event.
1) If you were to provide advice with AI what would you train it with.
I would train it first with the rule book, then I would give it the World Sailing Casebook, then I would give it any national authority Appeal books available online (RYA, US Sailing, Sail Canada). I would probably add written decisions from major regattas with experienced trusted juries (Olympics, Olympic Class World Championships). If I were training it to write decisions I would give it the Standard Wording document. I would consider what other resources were available online and use them as well. One of the reasons that we are behind on this is that not many articles are available online for AI to digest. In other fields there is a lot more published. I would consider some of the Sailing World rules articles, some of the articles on the rules in Dave Dellenbaugh's newsletter. If I were doing it myself I would add my own articles (https://rcyc.ca/Sailing/Racing/Know_Your_Rules)
2) What have you used it for
I have done a quick experiment with this. I don't know remember whether it was Claude or CoPilot. I gave it the RRS and the standard wording. I then gave it the facts founds by a fairly experienced protest committee and asked AI to write the conclusions. Some parts were correct and in fact did better than the original protest committee. Some parts need correction.
I have used AI to help me review NoR and SIs. For example I have asked "For major regatta's for xyz class do they modify such and such in the NoR" I have been returned NoRs with and without the modification. This can be very helpful.
I did try to use it to generate a pairing list for match racing. It didn't do well at all.
Wanting to do good for the sport, we instinctively want to help sailors. But as race officials we have to be fair to all competitors. There is a conflict, and especially on an international level.
When all competitors are able to access the same tools, without an unreasonable cost, an AI assisted help would be more fair to every sailor who can read. And it would certainly help non native English speaking sailors do better with protests and hearings.
All help is also needed for writing NoR and SIs for international events. They can always be simplyfied and standardised for the benefit of the sport.
Ps. My first ever comment in RRoS, and neither spell checked in American nor AI generated.
Ang