Meeting minutes
<wseltzer> https://
<wseltzer> see also https://
<wseltzer> (documentation of trials)
<wseltzer> https://
<wseltzer> https://
<wseltzer> https://
<wseltzer> https://
<wseltzer> present
Wendy: Welcome; take a look at agenda
… as usual, start with agenda-curation and introductions
… there was a question on list on FloC origin trial, documentation
… we had use cases pull request from AramZS
Agenda-curation, introductions
Wendy: and Robin Berjon would like to talk about a new proposal, GARUDA
… not sure we'll get to dashboard highlights, but let us know if there are things to put on agenda for future discussion
… Any other business or other agenda suggestions?
… Do we have any new participants who would like to introduce themselves?
… apologies if anyone is having trouble with W3C services; we have been having a few internal hosting issues
… hope on a path to resolving those
FLoC origin trial
Wendy: Arnaud, are you here?
Arnaud: yes
<Brendan_IAB_eyeo> Not another fiber cut in Brooklyn this time though, I hope? People will start asking questions of me!
Arnaud: thank you for making the time
… as many, we have started to collect FLoC IDs and make an opinion
… face problem of test being so small; 5%
… and users with Chrome beta version, doesn't represent a lot of people
… it constrains data and evaluation of proposal itself
… curious to know plan for population and geographical extensions, Europe, not only
… and see which browser was sending FloCs
… wondering if Yandex people are using?
… seems different from Chrome one
… so questions to Chrome and Yandex teams
Michael_Kleber
MichaelK: thank you for the question; unfortunately I don't have a substantive answer
… right now origin trial is only for a fraction of people using Chrome beta
… don't have new information
… we are getting some useful feedback from the origin trial
… and we will certainly announce it here when we intend to make any change
… I don't know anything about Yandex question, but would be interested
Arnaud: We see some insights coming up from FLoC origin trial; concern that data is not going to be representative
… We see that 90% of users changed their FloC
… kind of kills the advertising
… may be due to the specific population that test is being done on
… really double down on sample, and representation to really assess it
MichaelK: I don't disagree with what you said
… small scale will provide some useful info, but
… we do recalculate the FLoC based on browser history; how similar one week is to another is a question
… and use of FLoC algorithms
<AramZS> https://
MichaelK: some feedback is probably not very useful with a small scale population, and other types [of feedback] which are
Brendan, IAB Tech Lab: call attention to blog post
Wendy: Thank you
… also noticed Aram put a link into a Github issue
… along with question how do we want to incorporate documentation and trials
… wanted to open that for discussion as well
… if people want to add to this Github issue
<dmarti> (I am the main author of that CafeMedia piece, available to take any questions on it)
<alextcone> props to @dmarti
Wendy: or otherwise create additional resources
<AramZS> https://
AramZS: two things
… first is issue is 118
… how to incorporate documentation of results into the repo
… get to that in a sec
… I believe the idea is that FloCs would renew weekly in this current test
… they could stay stable
… I see Don [Marti] is on call and could answer questions for CafeMedia
… just call out concern with small scale of FLoC participants
… and some difficulty...with API response...as to whether people belong to a FLoC or not
… if others have issues interacting with the API
… Don's piece is great, about the data
… but also technical function
… I am going to move on to 118
… FLoC is first of the experiments starting to generate results
… not sure if people have specific ideas on how to document them
… The 'readme' is getting long
… could have sublists
… also interest in having proposals as stand-alone docs to put data against
… not sure what if anything people are interested in?
… If people have any ideas? Not sure if they do
Wendy: the light-weight approach is to post comments in an issue thread
… or if more gathering and curation, a separate document seems useful
… I have heard the concern that readme is getting long and difficult to navigate
Arnaud: Can we go back to first issue of the size of the sample?
… as many are integrating the API
… would be interesting for Google/FLoC folks to share how many FLOC IDs in US
… this would be telling for adtech vendors
… that championing any tests of FLoCs are over-selling it
… that is important to get those numbers to really set expectations
… I feel that we, some of us, know that the scale is very low, but some others feel that advertising is running on FLoC
… so maybe some numbers from the FLoC/Google team
Marshall: Thanks for feedback, we will take under advisement; appreciate the comment on the size
Use cases PR
Wendy: Any other comments on testing
… hearing none, let's think about how we show that the use cases and document them
… Aram, you had a suggestion on 117
AramZS: I can share my screen
<wseltzer> https://
AramZS: can everyone see my window?
Wendy: It's small
AramZS: can you see this?
[better]
… there was a bunch of discussion and maybe meeting before that on how use case doc is organized
… noted two factors; a lot of things to add, getting difficult to read. Lots of browser, user proposals, and comments getting incorporated into there
… another thing is way to sort or filter on how the use cases work
… I volunteered for a better way to do it
… here is my proposal
… did it with @
… puts it into a separate folder and creates a separate markdown doc
… use a bunch of new metedata header
… it can be multiple or single
… won't go into detail on how markdown works
… but takes this as metadata and uses it to generate a UI
… don't have info on how to build a process to take docs in
… but we need metadata to sort and filter on
… these are the various groups [walks through proposal]
… community; use cases addressed; current proposals; etc.
… here is what it looks like with a proposal
… and a template that describes what each doc should be, its structure and why it should work that way
… just a heads-up on what is there
… template explains why it should be structured that way
… in pull request, some notes about why structured the way it is
… and table with links to header down doc
… now links to these new use cases in details
… that is basically the proposal I have
… if it sounds interesting, we can move stuff into it
… Two questions for this group
… if we want to use this for a sortable doc, are these the metadata fields we want to use?
… is it the type of data we want to use?
… I'll lik the pull in irc
<wseltzer> [Scrolling through https://
<AramZS> https://
Wendy: this is a suggestion for making information more usable and more sortable; easier to add to
… as we have heard from previous calls
… and easier to filter by areas of interest, such as user or implementer
… if you haven't looked through those docs Aram linked, he is asking for feedback
ErikT: I don't have much helpful feedback
… but I like the format
… thanks for the work
James: thank you for raising markdown metadata
… seems like a reasonable way forward
… does W3C have any tools to look at?
… doesn't seem like a problem just this group is facing
… maybe next step could be W3C Team taking a look at analysis
… Seems like a more advanced way to go
AramZS: I think W3C has some templates
… not sure if they have rebuild tools
… would be great to hear
Wendy: I am not specifically familiar with these but will ask colleagues
<jrosewell> https://
RobinB: as far as I know, there isn't something exactly...but perhaps Respect and Bikeshed...could be helpful and happy to dig into it
<jrosewell> I was thinking about something extended from something like bikeshed. Thanks for reminder.
Aram: that would be useful; I was not familiar with bikeshed
<weiler> aram++ for asking for tooling help.
Aram: if there are build processes outside of W3C, let me know
… that would be useful to know as well; or suggest on the PR
Tess: with my terrible pedantic hat on...
… there is not a common markdown metadata format
… there are at least two
… and none is part of core specifications
… at risk...
… Yamel front matter
… and multimarkdown
… not sure what they call their metadata
… lots of metadata tooling
… @ does not support as far as I know
… there is a risk that some tooling may not support
… but many do support the Yaml...
… just point out that metadata in markdown is not a settled question
Aram: yes, that is a common concern
… one recognized by jekyll engine
… for any Github project to access
… would be shortest route for generating for markdown
… I love talking about metadata standards
… so let me know if people have preferences
Wendy: Thank you, Aram
<AramZS> @robin lol
Wendy: hope we can find a good way forward for integrating additional info
<joshua_koran> @Aram - THANKS - great idea - ideally we can also link to core definitions per Michael's Kitten Cluster proposal
Agenda-curation, introductions
Wendy: Next up
GARUDA
Wendy: we have Robin to share GARUDA; please go ahead
Robin Berjon, NYT: I have some slides
… can you see?
Wendy: yes
Robin: Hi everyone and thank you for having me here today to present this proposal
<wseltzer> https://
Robin: I promise a short presentation, about 10 slides
<AramZS> Also, @hober Also, I was not aware of the multimarkdown format! Thanks for bringing this up!
Robin: introduce a relatively complex topic, but stay at a high level
… this is a governance model for some chunk of what people are proposing here
… Governance of Ad Requests by a Union of Diverse Users is what GARUDA stands for
… several proposals rely on trusted server component
… one that's clearest is PARAKEET
… trusted server component in charge of guaranteeing an element and a few other useful pieces of functionality
… having a trusted server is quite useful
… can provide strong guarantees for privacy by serving as proxy for servers and networks
… can guarantee sovereignty and not leak data to/from third parties
… relatively easy to use existing adtech structure; plug into RTB system
… valuable given amount of infrastructure in existence
… value to do on the server, but also has to work on crap devices on bad networks
… practicalities of that can be problematic
… not have this take over too much
… but shared space, many stakeholders must come to consensus, so a bit of a forcing function
… What happens in advertising today is often unilateral
<AramZS> @weiler - I'm not familiar with that one. Is it best described here: http://
Robin: leads to unhealthy problems
… fraud, brand safety harm the ecosystem
… potential in future for this shared space to prove useful than just initial proposal
… So stepping back, what is this trusted server?
… who would ever trust a server?
… looking at ad ecosystem today
… users don't trust publishers
… publishers under revenue pressure
… intermediaries rushing to gain more data than users want to give
[missed]
… so trust is inherently difficult
… some proposals exist for using a trusted server
… let's trust publishers, intermediaries to run the server
… potential for SMC
… even though tech is great way to ensure enforcement
… they have to work
… not at level of maturity for secure multiparty computation to take over all this work
… tech means are self-enforcing
… if you get wrong, can be hard to correct
… said that making trusted server open source is a way to go
… but have to make sure software does right thing
… could rely on contracts
… but it's a 'pinky promise' thing
… could rely on regulations
… in long-term we need a tech solution
… and we cannot rely on browsers
… so options for trust are few and far apart
… How do we make trusted server and digital advertising possible?
… GARUDA...is a character that is king of the birds
… basically a governance model that establishes and organization
… that has a number of different responsibilities
… one is to define the technical standards for the trusted server
… to manage and operate and audit a worldwide network of trusted servers
… and ensure the overall health of the ecosystem
… Won't go into details of the governance
… rely on multi-stakeholder governance is supposed to look like
… one is a governance board
… another is the legal entity that runs day-to-day operations
… and is run by an exec director who decides day-to-day
… has a set of principles
… board has three constiuencies
… picked by how we can measure how much they rely on trusted server
… consider publishers and browsers plug in, and advertisers plus in
<wseltzer> https://
<wseltzer> https://
<wseltzer> https://
s/plug
… for each of constituencies, there is a threshold
… and once you achieve the threshold, you are eligible to join the entity
… voting structure
… guarantees some degree of consensys must be maintained
… and it's financed through a small tax on transactions
… details would need to be designed by group
… This is a web system, would build on web principles; the TAG principles come to mind
… there should be an advertising-specifc set of constituencies
… as there is putting folks into conflict
… today's systems puts intermediaries above publishers and users
… there is a logic to put users first
… principle of web and internet governance that is unassailable
… goal of advertising is to promote production
… publishers come next
… then advertisers who bring in money
… then intermediaries because others take precedence
… requirement of trust, so this org must guarantee privacy
… very important that publishers get to keep their data
… under the existing system the broadcasting of personal data
… means we are seeing attacks on high-quality publishers used to finance attacks to low-quality publishers
… this is not a viable way to run ecosytems
… publishers must maintain full sovereignty
… principles will be different
… universal access
… problem with existing approaches
… we need to learn from mistakes made in past
… problems of subsidizing universal access through a sustainable method
… there are issues
… this is not a perfect and final proposal
… there are two primary issues
… the buy side is not represented
… ways to work around that
… these people have significant market power and can vote with fee/wallets
… some representation, but consider some non-voting representation from World Federation of Advertisers
… mediated manner of representation
… sometimes problematic
… similar to buy-side, no way to establish a world-wide user polity
… I do have some other proposals outside of today's scope
… that could help to support this
… and zero doubts that this raises many more issues
<wseltzer> https://
Wendy: Thanks, Robin
… I added some links to the explainers while you were talking
… and look forward to link to your slides
James: Thank you, Robin
… looks like you have been wrestling with some of same issues as SWAM
s/SWAN
… you make statement about contracts
… contract law cannot be dismissed
… between entitties
… if contract failed in past, we should see what failed there
… we should think about advertisers
… Michael Donnelley from PRAM presented
… any input from that group?
Robin: Issue of contracts is broad and long
… contracts is not a good way to rectify power asymetries
… can site two examples of contractual failure in the digital advertising space recently
… one is entire set of contractual systems that are around ad choices
… the way being presented to end users is quite deceptive
… presented as a way to prevent behavioral profiling
… that works for half of cases
… but half of users believe they are opting out, who are actually being actively deceived
… there is this web of contracts that require people to support that, so it's hard for people to step out
… have to acknowledge the deceptiveness of the system
… no way to opt out of the contractual web, but they are also deceiving users
… Also there are asymetries of power
… The Times has been trying to be sole controllers of our audience data
… that is to say no third party can determine data processes on side without us
… despite being one of the largest publishers in world, we cannot get there
… we cannot decide all these issues
… significant market power
… intermediary capture
… system without multistakeholder arbitration
… stakeholders with power, not as one entity, but as aligned entities, get to call the shots
… that is limit of contractual approach
… I have not spoken with PRAM
… but have spoken with several buyside companies
… will invite them to the conversation
… So far, there is definitely interest
… buying power might be significant for them
… bringing this to public light to obtain input
Mallory: Centre of Democracy and Technology: clarify the kind of advertising you are talkinga bout
… impression advertising
… cannot encompass all
[scribe has to step away]
<wseltzer> mallory: people, beyond users of the web
<jrosewell> Re Robin's comment on "dark" UI patterns - https://
<wseltzer> ... need to consider the human rights or public interest, not just "consumer" or "user" perspective
Robin: Thanks, I don't see those questions as pedantic at all; that's the detail we have to get right
… what you are saying is in line with discussions at @
… so this is in line with feedback received and aligns with issues that I would like to solve
… to try to get to that
… in terms of kinds of advertising, that depends essentially on that this solution plugs into
… meant to be a complement to existing advertising options
… designed primarily with PARAKEET in mind
… it's not behavioral tracking across the internet
… don't have a crisp answer; meant to be respectful of users
… how to provide support and to take into account a broad set of people without having a polity for them
… If I understand you correctly, picking a set of principles that aligns with that ideal, such as human rights framework
… that is interesting
… if we pick those valuable principles, but not have representation, we fall into concern of not having them check if true
… try to get some representation of people who do this for a living
… no decisions or anything firm yet
… what I am worried about less about defining principles, but about how to keep ourselves honest
… maybe public interest orgs can check our work and make sure we are doing the right thing
… but this org will be financed and might not provide financial support to do checking that things are right
… I know this is not a perfect answer, but hopefully enough to show a path forward
Mallory: Two things, accountability and representation, and not entangle the two as they are both important
Robin: maybe get some orgs as observers, but worry that we have five American orgs, two European, one Asian and that would not be representative
James: quick follow-on questions
… you talked about power asymmetry
… just a thought
… I've been looking at CMA and ICO
… have not spoke to UN
… was it the UN Internet forum?
<jrosewell> https://
<mallory> The problem is that intermediaries (the fourth entity) see themselves as the mediator of the three. And based on Robin's schema, intermediaries are at the bottom.
Robin: Taking points in order
… there is value to have tension with principles, and representation from consituencies
… but you always end of up with undecidable problems
… and that tends to be provided by a priority of constituencies
… I take point that some may try to weaponize the priority of constituencies
… why need for power framework
… this has gone before CMA and other regulators
<jrosewell> Agree about "weaponize" constituencies. Everyone attempts to speak for people.
Robin: I think we need priority of constituencies to address issues
… In terms of UN, this was a discussion with data fellows tehre
s/there
… I don't believe they report to Internet Forum
… report to some innovation thingy
… looking at issues of data colonialsims
… making sure there is equitable use of data
… don't want to represent that anyone said what they would support on behalf of the UN
<mallory> Thanks and goodbye everyone-- need to join another call.
Erik Anderson: regarding PARAKEET
… Microsoft is happy to be in conversation to express concerns
Robin: in terms of next steps, anyone who wishes to engage, happy to set up follow up calls and discussions
Wendy: Thanks, Robin
… you are thinking big about institutional design and big governance questions
… will add to the readme and repository where people can file issues and comments
Robin: thank you
Wendy: Thanks for the presentation
<wseltzer> [adjourned]
Wendy: please keep using agenda plus tags to queue up other items for discussion, and see you next week
[adjourned]