w3c
Recusal from FO handling
Fuzzy criteria for recusal are problematic: individual with only tangential involvement in the matter, or even with no real involvement at all in the matter, could be pressured into recursing themselves. Conversely, over-assertive individuals may fail to recuse themselves even when it would be appropriate. We may instead want to make a strict and non ambiguous criteria. It is also unclear how useful it is to require recusal from the person raising the objection, since that can easily be circumvented by asking someone else to file the objection on one’s behalf. For example, recusal could be required of (and only of): a Chair who issued the Chair Decision or Group Decision being objected to, and the CEO for objections against their own decisions and those of the Team.
See https://www.w3.org/Consortium/Process/Drafts/director-free/#addressing-fo
No one should be both a party and an adjudicator. We don't necessarily need to define further.
@wseltzer that's a really good general rule. From the tone of the issue text I sense that there is a lingering question: "who is a party?"
The decisions, however, affect the Web so at some level everyone has at least some interest in the difference of opinion. The parties shouldn't be excluded from the conversation, only from the final process by which the committee reaches its decision.
So, I think we converged.
No one shall be both a party and an adjudicator. The person making either the decision being objected to or appealed, and the person making the objection or appeal, must recuse themselves. Recusal means that they do not form part of the consensus or voting body that issues the decision; however, they may take part in deliberations leading to the decision.
@dwsinger wrote
No one shall be both a party and an adjudicator.
This makes sense, except perhaps where the decision being contested is a collective one and the decision-making body would normally be a substantial part of the council.
My suggestion for that is that the bodies do not recuse themselves. They make decisions by consensus based on the available information. Presumably an appeal to the council will bring some new information (perhaps only the unanticipated depth of objection, but that is information) which means the council reconsiders. And each body is less than half the council.
There is a bias toward the original decision in this case - much as I suggest there should be in cases where there isn't a strong (supermajority) agreement on the council to overturn a decision. I think that is probably OK...
ah, right, adjusting for clarity for that case:
No one shall be both a party and an adjudicator. The person making either the decision being objected to or appealed, and the person making the objection or appeal, must recuse themselves. Recusal means that they do not form part of the consensus or voting body that issues the decision; however, they may take part in deliberations leading to the decision. Recusal is individual; when the decision in question was made by the TAG or AB acting as a body, the entire TAG or AB is not expected or required to recuse itself.
That proposed paragraph has multiple parts. Let's go over them individually. For reference, here is the director-free branch’s current text.
Part I: Required Recusals
Proposed wording:
No one shall be both a party and an adjudicator. The person making either the decision being objected to or appealed, and the person making the objection or appeal, must recuse themselves.
Current draft's wording:
Participants in the W3C Council must recuse themselves if they are the person who issued the decision being objected to, or if they are the person who registered the Formal Objection.
The proposal like an editorial improvement, so suggest we adopt it.
Part II: Defining Recusal
Proposed wording:
Recusal means that they do not form part of the consensus or voting body that issues the decision; however, they may take part in deliberations leading to the decision.
Current draft's wording:
Any person who has recused themself does not take part in its deliberations, determination of consensus, or votes on this matter. The W3C Council may still solicit and hear their testimony, as they can of anyone else in the W3C community.
The current draft's definition of recusal also blocks participation in the Council deliberations (but allows the council to solicit their testimony), which seems more appropriate particularly for a consensus-based decision. We suggest keeping the current text.
Part III: Clarifying group vs individual recusal:
Proposed wording:
Recusal is individual; when the decision in question was made by the TAG or AB acting as a body, the entire TAG or AB is not expected or required to recuse itself.
This point is not considered in the current draft; we would support adding it.
Part IV: Optional Recusals
In https://w3c.github.io/w3process/director-free/#WGArchiveMinorityViews, the director-free branch of the process also says that:
Participants may also [=recuse=] themselves for any reason.
As observed during the council experiment, this can easily lead to confusion about the purpose of recusals and the difference between recusing and being absent or busy. @fantasai and I suggest the replacing that with:
Participants <em rfc=2119>should</em> [=recuse=]
if they believe their participation would compromise the integrity of the Council decision.
However, recusal is disqualification from participation,
not abstention,
and <em class=rfc2119>should not</em> be used
to excuse an absence of participation.
-- @frivoal and @fantasai
Thanks, I agree on the last point. Recusal has a very specific meaning, (e.g. Cambridge) "the fact of a judge or a member of a jury not being involved in a trial because they have a special interest in its final result". We need to make this clear; people who are recused are excluded for a reason, not merely choosing or unable to take part. A lack of time, a lack of interest, a lack of expertise: these are not recusals, but a reason to abstain or stay silent. Indeed, if there is a vote, members may abstain or even be absent: that it not recusal. Given the confusion we have had, I think we need to add the definition, and explain that recused members have no influence and little or no visibility into the deliberations and decisions.
Thanks Florian, for refocusing on recusals.
I have a different set of observations from the experiment. I had planned to make those observations in the experiment post-mortem, but as long as we are going through the text, I might as well make those observations here.
Above, we required Council members to recuse themselves if they were the decider (of what is being objected to) or if they were the objector.
Looking at the experiment, I don't think that is strong enough. I'm talking about several people who participated in the Council and did not self-recuse and their decision is in accordance with the above guidance.
However, these people work for organizations who have business interests in the results. A priori, it is not obvious why they have less bias than the objector. So, imho, there were additional people who should have recused. I don't have a particular proposal at this stage - the levels of business interest vary - so we need discussion as to what level requires recusal.
I think we all (or most of us) have a business interest in the health of the world-wide-web, so I don't think that we'll easily find a bright line.
The reason to have a 'may recuse' is precisely for this case; if you feel, or someone suggest to you, that you appear to have too much interest in one side, then recuse. We'll never succeed in delineating all the cases; we're looking to people for an honest assessment of the propriety of being involved in the decision. 'My spouse filed the original complaint; I recuse myself as otherwise it might be seen as my support of my spouse.' for example.
That's why we set the automatic obligatory recusals so narrow. Perhaps we should explain this more around the 'may recuse'.
That's why we set the automatic obligatory recusals so narrow. Perhaps we should explain this more around the 'may recuse'.
Just be clear, I think we have set the automatic obligatory recusals to be too narrow. Objectors deserve better clarity that the Council Members ruling on their objections have no conflict of interest.
@jeffjaffe I disagree, or at least doubt that there is any other easily-defined 'bright line'. Almost everyone has some interest in the outcome. It's not a yes/no, it's a question of degree. The question is whether those factors rise to the level that they must be excluded.
Perhaps we can strengthen the 'request to recuse' part, and explicitly state that people who others perceive have a COI should recuse when that is brought to their attention. But I don't want to provide a weapon for getting rid of people perceived as 'troublemakers', either.
The Director has or had a COI when ruling on matters that affected his personal area of research or startup, for example.
The Director has or had a COI when ruling on matters that affected his personal area of research or startup, for example.
It was appropriate for the Founding Director to rule even when he had a modest COI. Given his stature - the advantages overwhelmed the disadvantages.
I don't think that applies as clearly as we create a new system. Quite the contrary, our new system needs to give confidence that it is as unbiased as the Director has been.
I don't have a proposal at this stage. I hope to discuss at the experiment debrief.
To me, there's broadly two kinds of situations to consider:
- People who have some interest in the case, and whose position is likely to be biased. To one degree or another, I think this is everyone. Some are more closely involved than others, but we all care about the web, and so we all care about the outcomes. I don't think we can find a clear line between those who are too involved and should recuse, and those who are removed enough that they can be sufficiently impartial. I think we overcome this by being a large diverse group, with a fairly high level of trust thanks to being elected, and talking it through, not through recusals.
- People whose actions were are discussing, and whose mere presence is likely do distort the discussion. If I file a formal objection, and other members think my objection is abusive and I am a troll, they may have somewhat of a hard time expressing that if I am there for the discussion. Courtesy, respect for the CEPC, and the general awkwardness of criticizing people to their face, especially without having a chance to check with other people if they feel the same as you do, would stifle the discussion. I think recusal is appropriate for this case.
Also, if the bar for recusal is both somewhat fuzzy and somewhat low, I fear that this will bias against people with high integrity, which seems bad.
At the experiment debrief I pointed out that potential Council Members who work for companies who have a strong business interest in the results of the objection should be required to recuse.
We could include such a comment in the formal process, but that would then cause uncertainty about the definition of "strong business interest".
It would be worthwhile if we can make this more objective and less subjective by actually spelling out a defintion of strong business interest.
I am somewhat doubtful we can succeed in making a clear enough bright-line definition of "business interest". However, I do think we could consider even more subtle, layered approaches than the two-level we have now: mandatory (must recuse) and optional (may recuse).
For example, we could make it explicit and on the record that anyone, in or outside the council, could suggest that someone recuse. It can't be forced recusal, or it becomes a weapon to get rid of people we don't like.
We could allow 'both sides' some "peremptory dismissals" but I fear again for mis-use, and though the person filing the FO is usually clearly "one side", it's not always obvious who the "other side" is (e.g. in this case, where a charter was drafted by the team working with the chairs and other leaders and supported by people who want to advance some work items).
If we had an organizational pre-meeting (which seems to be a good idea), we could allow the council to consider whether someone needs to recused, and decide or vote on it.
For what it's worth, I think having a formal mention made of on-the-record requests-to-recuse (the first idea above) is probably enough. Maybe.
Overall, I don't think recusal is the best too here to push back against "would be biased" for various degrees of bias. We counter that by having a large and diverse group, which on top of that derives legitimacy through election. I think recusal is more useful and more easily applicable to case of "would stifle the discussion and prevent the council from discussing the maters fully".
I do think mandatory disclosures of (potential) conflicts of interest is a good idea.
@frivoal the problem with countering "bias" with "a large and diverse group" is that this large group still can be homogeneous with respect to certain objectors. As you point out, this large group derives its legitimacy through election of members, but web stakeholder's interests are broader than Member interests.
Since we have a "large and diverse" group to select from, what is the harm in having recusal of those that have strong business interests in the results?
The harm is that you still have not defined who judges what constitutes "strong" so it's subjective.
Is the "experiment debrief" available, or was this confidential to AB members or to W3C members?
Is it true that this issue is only relevant if the suggestion of handling formal objections through a combination of AB and TAG members is adopted?
I think it would be nice to be able to recuse all people who have a strong interest in a topic, but I doubt that it is practically compatible with having an appropriate diverse set of people to judge an issue.
- defining bright-line rules is at least as hard as @dwsinger suggests (I don't see even a first stab at it in the discussion so far, and that's telling to me)
- we are basically talking about people who are volunteered by their employer to work in the public interest. There are discussions elsewhere of why that's not always a strong motivation to do more work.
- diversity will increase the chances of someone being able to act as an effective advocate for a position, in order to ensure it is given due consideration. This is true even for the case that I think is inevitable, where there is nobody on the council who personally supports the substance of one side of an objection.
The risk in recusing too many people is that decisions are produced by a lopsided group. W3C has the trust of the community that it generally works for the public interest in the Web Platform, and that on balance it does good things. If the decisions made in future are perceived as lopsided, that reputation will be damaged, and with it W3C's ability to attract the participation it needs. Likewise, if it adopts processes that are seen as arbitrary and capricious, so a recusal policy itself might be the damaging thing.
The harm is that you still have not defined who judges what constitutes "strong" so it's subjective.
As you know we discussed some of the objective criteria at the debrief.
In my opinion, if a company is implementing the technology that is being objected to, employees of that company have the appearance of bias (even though in the examples I know (real TAG/AB people) the people are not biased). Accordingly, I don't think that such employees (even if they are wearing their TAG or AB hats) should be on the decision panel.
I'm looking at it from the perspective of the objector. I believe that the objector would reasonably believe that they have lost those votes even before formal objection analysis starts.
In my opinion, if a company is implementing the technology that is being objected to, employees of that company have the appearance of bias (even though in the examples I know (real TAG/AB people) the people are not biased). Accordingly, I don't think that such employees (even if they are wearing their TAG or AB hats) should be on the decision panel.
Particularly given how widely we want our techs adopted, "anyone who's employer is implementing the tech" is too low of a bar.
I hope that someone who feels that employer pressure would lead to them being unable to vote their conscience would recuse, no matter their or their company's stance. ("Dear employer, I serve on the TAG - our company does not. If you tell me how to vote or threaten me with retaliation, I must recuse - even if I agree with you.") Do we (instead) need some words encouraging the independence of TAG and AB members?
I was directed here to assist with the current language in:
https://www.w3.org/2021/05/W3C_Council_Guide.html
The W3C Council determines who will rule on this objection:
-
If any of those individuals have a formal Conflict of Interest related to this case (e.g. they are also the Chair of the Working Group that made the decision that is being objected to), they MUST recuse themselves from the Council for this case.
-
If they are concerned for whatever reason that they cannot be objective about the case, they MUST recuse themselves.
-
If they do not think that they can spend the time required to provide a timely ruling, they SHOULD recuse themselves.
- Further elaboration is important here. Every person serving on the Council is expected to be sufficiently knowledgeable and thorough that they feel comfortable with the decision of the W3C Council. There is no value of having a large Council of participants that defer to a smaller number of more knowledgeable members on the Council. If that were to happen there would be a non-transparent situation where the Council decision appears to be made by a large grouping, but in fact is decided by a few. Better to recuse than to "go along with the majority".
I think the potentially problematic one is:
If they are concerned for whatever reason that they cannot be objective about the case, they MUST recuse themselves.
Some folks probably trust others to be objective regardless of the employer stance, whereas others might feel it is impossible to separate the positions held by an individual from the material impact a web standard may have on their employer.
I would tend to not trust an individual to decide if that is the case or not, and as such I am not in favor of voluntary recusals, but I could be convinced.
To protect the W3C Process, even the appearance of bias should be avoided.
I was directed here to assist with the current language in:
https://www.w3.org/2021/05/W3C_Council_Guide.html
Thank you; I fear you will need to read the comments in this issue, as well...
@dwsinger in case its not clear from my previous post:
In my opinion, if a company is implementing the technology that is being objected to, employees of that company have the appearance of bias (even though in the examples I know (real TAG/AB people) the people are not biased). Accordingly, I don't think that such employees (even if they are wearing their TAG or AB hats) should be on the decision panel.
I'm looking at it from the perspective of the objector. I believe that the objector would reasonably believe that they have lost those votes even before formal objection analysis starts.
https://github.com/w3c/w3process/issues/278#issuecomment-767600289
This is from the perspective of the objector, the same holds for working group advocates or members.... for example, if apple objects to web nfc, should they also adjudicate the formal objection? Since apple does not plan to implement web nfc, no conflict of interest right?
What about mozilla objecting to trusted types in web app sec wg charter? they don't plan to implement it, so perhaps they have no conflict of interest and should be allowed to adjudicate if trusted types are covered by the next charter.
Seems like the best solution here is to say:
- If we don't trust a TAG / AB member who is an employee of an organization in favor of the standard to be impartial.
- Then we don't trust a TAG / AB member who is an employee of an organization opposed to the standard to be impartial.
And my original point was, are voluntary recusals enough?... IMO, no... its safer for W3C to just handle this stuff consistently, for both objectors and advocates... we should have a process which is very easy to defend, and voluntary recusals for objectors implies voluntary recusals for advocates....
Unless I am missing something, there is no specific solution for recusal of objectors or advocates currently, and it is possible for an advocate or objector to be an adjudicator if they think they can handle the appearance of bias.
@OR13 I'm not entirely sure what you're proposing for non-voluntary recusals. Is it "all potential Council members whose Member employer voted for or objected to (i.e., any response but 'abstain') the question at hand must recuse"?
I'll note what I put in email - this unfortunately incentivizes not voting for any Member who employs someone who is elected to the TAG or AB.
There are insightful comments on this topic in an email thread on the AB repo, which I would prefer not to replicate here in its entirety.
In that thread I suggested this:
- the set of formal must-recuse is very small: the person (individual) filing the FO, and if it’s about a decision made by a person, the person who made that decision (e.g. a chair decision)
- anyone can self-identify “my presence on the council would taint the decision, for a reason that must be stated, and I recuse myself”
- anyone can suggest to the team shepherd that the shepherd consider whether someone should be asked to recuse themselves: 3.1. the team shepherd can ask anyone to recuse themselves, weighing the input received 3.2. if the person doesn’t recuse themselves, or the shepherd declines to ask them to, and anyone (including the shepherd) thinks it’s a serious problem, the council will hold a simple majority vote before it takes any other business, on whether to recuse that person
I suspect we’d rarely get to 3.2
I'm still unconvinced by any recusal proposal - the set of people without an interest in the issue seems approximately the set of people least likely to inform a good technical decision on the merits of a case...
That someone has an interest in the case does not mean their interest is not legitimate. A case where someone is so conflicted they promote a dishonest argument that would actually pervert the desired function of the decision-making process seems vanishingly unlikely, and especially unlikely to pervert a consensus among a diverse W3C council engaging in clearly recorded discussion as a precondition to reaching a decision.
Enabling a respectful way to suggest the interests at play, as part of the record of the discussions that lead to a conclusion does seem a hard, but worthwhile aim, in that it helps people understand the context of the discussion, and so helps motivate honest discussion and conclusions.
@cwilso I agree with what you put in your email, specifically this part:
this unfortunately incentivizes not voting for any Member who employs someone who is elected to the TAG or AB.
I also agree with @chaals
the set of people without an interest in the issue seems approximately the set of people least likely to inform a good technical decision on the merits of a case...
I think @dwsinger 's suggestion here is excellent: https://github.com/w3c/w3process/issues/278#issuecomment-940496920
I don't know enough about the "team shepherd" to know if they are trust worthy, or a single individual or a group of W3C Staff... but they seem to be providing a valuable role in reducing the size of "mandatory recusal" which would seem to be required to get good technical feedback.