We should open to the mute and other speech disorders

Faces are unique and recognizable, voices aren’t. Although they are unique like faces, we haven’t evolved to recognize them perfectly. Machines can, but current vouching is done by humans. I propose we replace voice text only for a selection of options, adding options for those who can’t comply with one:

-Unique voice read-out (actual situation)
or
-Unique gestures (define a set of face gestures)
or
-Unique corporal movements (define a set of movements, ie.: walking is unique for each person: Why you are identifiable after just a few steps | New Scientist)

6 Likes

Yes, those options look valid. Also they could make hand gestures of their wallet address for example (a very very hard thing to do today).

1 Like

I think we should try to accommodate for those particular cases but should also be careful not to open a breach in POH.
The problem of allowing different ways to register is that attacks could try to register multiple times using the multiple ways. Maybe asking a doctor note (including a way to contact the doctor that people could use to try to catch “fake doctor notes”) and limiting the proportion of mute applications to double the proportion of mute individuals would lead to a nice balance between security and accessibility.

Agree, but please detail how multiple entries with multiple ways is different to multiple entries with the same way. It’s the same character in both, so wouldn’t that be a duplicate entry still? I Don’t see the advantage for attacks.

Like if one person make a submission with voice and without voice, people running voice-recognition algos would not be able to match those submissions as being duplicates. Normally it could be matched via other methods but that’s one less way to detect which is gone.

Easily solvable with other types of algos, like DeepLabCut — The Mathis Lab of Adaptive Motor Control

1 Like

I had a wonderful, unique life experience several years ago when I interacted with and made friends with some members of the deaf community, some of whom attend Gallaudet university, located in Washington DC, the only liberal arts college for the ASL community in the world. I actually have a couple deaf and hard-of-hearing friends who could be interested in POH. So there is some demand and interest. The hard-of-hearing friends and deaf people who weren’t born deaf can usually vocalize, and could probably attempt the phrase “I certify that I am a …” but the born deaf are usually mute.

The deaf community is very tight-knit. There isn’t a lot of overlap between the hearing and deaf communities. It’s pretty clear that a network of deaf people who vouch for the deafness/muteness of other deaf/mute people would have high connectivity. Perhaps to ensure robustness, proof of muteness would require many vouches (5+? 10+?)?

Then only those registered in the POD / POM (proof of deaf / proof of mute) registry could register for POH without the voice requirement.

A genisis set of deaf/mute people who vouch for other peoples deaf/muteness? Is there even enough critical mass interests amongst deaf/mute people to start such a network? Is this pulling the cart before the horse? Or is this a chicken and egg problem where deaf/mute people are routinely excluded from hearing communities including the blockchain and ethereum communities. If deaf/mute people are always excluded (in many aspects of society), then they will never find out about the greatest innovation (blockchain) in society since the internet and printing press.

Maybe POH needs a more comprehensive solution to all humans who cannot participate in POH for one reason or another (partial face paralysis due to stroke, homeless and cannot manage private keys, children, other handicaps). Perhaps POH could be inclusive in $ubi issuance, but more selective in the $vote distribution as commented on the cases of children (danger of perverse incentives) and homeless with guardians/care takers/ ‘do gooder’ organizations

2 Likes

Having a discussion in poh governance I thought about a possible solution. Let me know if this could work somehow. The idea is to create sub-courts of jurors in the Humanity Court for each specific case, on the free will of whomever thinks they have a special case. Let’s think some person wants to register someone that uses LIBRA (brazilian standard sign language).

  1. For the first set of jurors in the Humanity/LIBRA subcourt, a LIBRA juror self candidates to offer their expertise in that language, by creating a dispute in the General Court. As evidence, they present their credentials showing how they acquired such skills. If passed, they stake the specified amount of PNK and are added to that subcourt. Do this a couple of times to get a reasonable amount of jurors.
  2. Whenever a LIBRA-speaking person wants to register, they start the whole process of submission, and in the video or somewhere in the name or other field they specify they require a LIBRA court. Once in the pending registration phase they are challenged by default.
  3. By some mechanism (yet to be thought of) the subcourt is triggered and that disputes goes there. The specialists decide that they fulfill the requirements or not.

This is just a sketch of an idea, there would still be things to be sorted but it might work. I am really interested to hear from @clesaege @fnanni or anyone from the Kleros side.

This would be quite complicated to implement (would need a new version of POH).
Maybe a way to deal with this could be to have people bring a doctor note asking for an exemption of talk and we’d limit the % of people with exemption to a value slightly above the % of people with an exemption enabling disability.

It could be done with the current contract, the exemption would just be metadata and could be calculated from them (to see if there are still some exemption slots remaining).

The idea of having a maximum % of exemption slots is that even if someone were to be able to abuse it, this couldn’t corrupt more than a specific % of the set registered humans and it’d raise some alarms (i.e. if the rate of disabled people registering is significantly higher in PoH than the general population, that would probably mean that there has been some exploit).

1 Like

Ok, usually people with disabilities tend to have some government issued certificate (to access special accomodations). This could be legitimate enough as evidence (or maybe more than one proof would be required.

Caps could be obtained from census data (if available). A national level would be enough, to set them.

So which steps could make this work? Where should the caps be set and how and by whom?

1 Like

For cases in which the human cannot speak, it cannot be done with someone who speaks for that person, someone who is already registered and says something like: I certify that he / she is a real human etc …?

This could work for babies and kids that do not speak, somehow. In the case of adults, there could be a risk of unconsenting people being registered by “farmers”.

The right tool for this kind of idea would probably be Linguo secured by a sign language subcourt. On profile submission to POH, users could choose to use sign language. If they do, they would first submit the video to Linguo and automatically self-assign the translation task to themselves and post the correct english translation, which should be the registration phrase described in the POH policy. The Linguo-certified video could then be submitted to POH.

The problem I see with this is: (1) UX for something like this is a bit hard right now and (2) sign language is not universal. As far as I know, there exists an esperanto-like sign language but is not widely used. However, taking into account that it’s just a phrase that people with speech disorders would have to learn, maybe it is worth considering it.

The same concept applies to any language.

@clesaege what do you think about this?

3 Likes

International Sign is actually very widely used.
Rather than the “Esperanto” in terms of adaptation, it would be the “English” of sign language, though it is indeed somehow built as Esperanto, with 50% of it coming from ASL (American Sign Language).

Why are we making it harder for people with disabilities to join?

This is a common mistake where we disable people in a systemic way.

Having deaf friends (I can sign some ASL as well) I think this is very bad.
First: someone would have to show a document with their full names/identity, which currently nobody has to do in PoH and everyone could have a fake name.
Second: these documents are indeed a thing in rich and Middle-income countries, but not in most of the world.

Can we find a way that won’t add an extra layer of difficulty to the process?
Can’t we have them make a video showing a paper/text where they write the phrase and sign it with their name on the platform?

1 Like

The main problem with this approach for me is that it makes it easier to “farm” profiles from non-english-speaking people without the current barrier of convincing them to speak a phrase. You could also do the same with kids and babies.

I’m thinking out loud, but maybe PoH could be the “layer 1” identity protocol, and other accommodations could be made with "layer 2"s that allow for different rules.

For example, maybe someone could coordinate the creation of a “proof of muteness” and/or any other disabilities, a parallel protocol that would work as an universal, cross-government issued certificate of a given disability. Maybe working on this modularly would give better results, since the specific needs of each population could be targeted more directly by its community, and all the required accommodations could be made.

Then, after this protocol is created, from the PoH side we could then allow the list of profiles in that second registry to be exempt of a rule that it’s difficult for them to comply. For example, mute people could be required to hold a sign with the phrase written on it. We would just need to specify that in the the rules in the Policy Document, and provide to challengers/jurors some flag in the UI for these profiles, and as a backup, where to look up the list.

Also, the cap mentioned by Clément could work as a backstop for abuses of the “layer 2” protocols that accommodate different rules.

Later, we could then adapt PoH v2 to integrate these concepts more natively if necessary.

My thoughts on the subject:

  • The security of the PoH registry is of the utmost importance since PoH is used to distribute voting power. It is important to be inclusive of everyone including deaf people but this must not come at the cost of the registry’s security if at all avoidable. As a reminder, we already have a farmer in control of over a hundred profiles so we should probably already thinking of tightening security.

  • Requiring government certificates or even a doctor’s note goes against the spirit of PoH: it reveals more information than necessary about applicants, can lead to censorship from government bodies, and not everyone in the world can gain access to such documents in the first place. Not to mention that a doctor’s note would be trivial to forge so it does not seem to me that it grants much security at all.

  • So the question is: if we remove the requirement of speech for deaf people, what other security parameters can we tune up? It has been suggested to increase the video requirements by requiring specific facial expressions or full body movement. These are interesting ideas which I think could in time be required of everyone as deepfake technology improves but they do not solve the problem of making sure the participant fully understands what they are agreeing to. Instead, I feel like we would be better off tuning up the social graph aspect of PoH’s security for deaf participants. Here’s a proposal that can be implemented without modifying the PoH smart contract:

Proposal (very rough draft):

  • Participants who are unable to speak due to a medical condition may submit an application without saying the phrase “I certify …” out loud, but they must instead physically appear alongside a registered PoH user (henceforth, the helper) who must say the phrase: “I certify that the applicant appearing next to me is a real human and not already registered in this registry. I certify that they are unable to speak, that they understand the purpose of this video, and are registering for their own sake.”.
  • The helper may only provide this service for a single profile. A person registered through this method can (obviously) not provide such a service. (rationale: This should severely limit the extent to which a single person or even a small group could abuse this registration method.)
  • The sign must display the helper’s PoH address alongside the applicant’s address, in this format: “Applicant: [applicant address] Helper: [helper address]”.
  • If it can be proven that the applicant is capable of speech, the profile is invalid and is to be rejected or deleted.
  • The applicant must be vouched by someone other than the helper (this is a way of ensuring two “vouches” for deaf persons without having to modify the smart contract with this special case).
4 Likes
  • Doctor notes even if trivial to forge, can be used as evidence (if one doctor is giving tons of them or that this doctor cannot be joined, is not on the national registry etc).
  • Having someone different come with the user could work. Making him speak offers some protection against profile farming by asking people in the street.
  • Actually we could require that the person vouching comes in the video, this would make it easier to verify that one person doesn’t speak for multiple persons (as vouches are easily tracked, while a random person without on chain link is harder).
  • I still feel like it would be difficult to check a doctor’s credentials. How hard is it to forge a diploma? Maybe not so easy in developed countries? Maybe. But what about other countries. There’s also the fact that not everyone in the world necessarily even has access to a proper doctor. And in corrupt countries, would anyone even care if a doctor gave out false certificates in exchange for bribes?
  • Another big protection from profile farming in my proposal is that PoH members can only help a single mute person each.
  • The reason I did not propose that the helper should be a voucher is that this would be a vector for attack with the current PoH smart contract since the order of events is the following: create profile → get vouch → pass to pending. This would allow a so-called vouchallenger to lie in wait and, as soon as a mute profile is created, to vouch them and challenge them before the helper has time to vouch them. This would not be an issue in a future iteration where a profile could only be passed to pending with authorization from the applicant. EDIT: Note that in my proposal, the helper would still have to be a member of the PoH registry and would have to display their address. I imagine it would not be too difficult to update the UI so that the helper’s address is added to the json metadata stored on IPFS with every submission. The data would not be on-chain but in an IPFS file linked-to from an on-chain event, so almost as good.
1 Like

What if we just allow, just for these cases, to be assisted by another human that can certify this individual is a real human and has not been registered in the registry already?

I guess that wouldn’t generate any breach as it would require to have 2 humans in the video, instead of one. The Human that can’t speak should be the one holding his/hers address.

If we wanted to be even more secure, we could require the Human assisting to be already registered in PoH. If someone can assist you to complete your registry, for sure he can registry himself.

This idea would also be useful to human cases that also cant make unique gestures (f.e. a baby or really young human).

Of course solution is not perfect, but seems to be really similar to what we have today, and would be a solution for most cases that are not available to register as for today, although they should be able to.

Luis.

2 Likes