Ideally, some consensus-building process should occur before committing taxonomic changes, especially those that affect a large number of identifications. This has been discussed for example in this forum topic and in the comments on this iNaturalist blog post. However, both of those discussions have broader scopes, and I think it’s time for a more focused conversation on what the protocol for reaching consensus should look like and how we assess whether or not consensus has been reached.
The curator guide is mostly silent on what consensus-building should look like. Here is what it does say:
Try to @ mention others to review your changes for potential errors and to discuss whether or not they’re appropriate. This is especially important if you’re changing a taxon based on a regional authority and it has observations outside that region. Curators have a lot of power to act unilaterally because sometimes it’s just hard or impossible to get others to vet your work, but we (the site staff) would prefer a more collaborative process whenever possible.
For vascular plants, taxon changes now require 5 votes (not including by the curator who drafts the change) before they can be committed. Formerly, a one-month waiting period was required between flagging a plant for a taxon change and committing the taxon change, so that the community had time to weigh in on changes. Both of these policies signal that staff want to restructure the taxon change process to incorporate/require collaboration building.
In spite of the above, here are some problems I’m seeing:
Plant taxon changes are still moving through the 5-vote process without any serious discussion. This week, a split disrupting tens of thousands of IDs was committed without recent discussion, even though a years-old discussion favored holding off on the split. The curator who drafted and submitted the change stated that she intended to build consensus during the 5-vote process (evidenced by tagging identifiers), and the curator who committed the change after it very rapidly received 5 votes stated that he interpreted the 5 votes as sufficient evidence of consensus. Both curators acted in good faith, but the end result was a change that wound up eliciting significant blowback. When in the taxon change process are we supposed to be soliciting feedback? How much feedback is required, and how much time should be allowed?
In other examples, a single dissenting voice completely stalls progress on a taxon change. Certainly we should be careful about changes that are not unanimous, and a single voice on a taxon flag probably represents the thoughts of many people who remain silent. That said, consensus is generally not the same as unanimity. We already have a threshold for how many IDs is takes to allow the community ID to advance in the presence of “maverick” IDs. How are we measuring consensus when assessing whether to move forward with a taxon change? How many dissenting voices equal lack of consensus? What do we do when consensus cannot be reached?
Other projects, including Wikipedia have explicit guidelines that address the questions above. I would really like to see the same kinds of best practices outlined explicitly in the curator guide, at least for plant curation, which seems to get more acrimonious every year. In the meantime, I would really like to hear other curators’ ideas about the questions I posed above and related concerns.
Good topic. The bar needs to be higher than was in place for the Rubus ursinus split that @holyegg mentioned. For example, a one-month waiting period would have helped.
“…and the curator who committed the change after it very rapidly received 5 votes stated that he interpreted the 5 votes as sufficient evidence of consensus.”
This curator is being, I think, purposely obtuse. The vote tally itself is clearly not evidence of consensus in the context of an ongoing discussion — especially when the votes accumulated rapidly — however you define it.
I posted this previously on another thread and stand by it.
For vascular plants, I wonder if a series of required fields/check-boxes would help. Basically these would be a checklist of questions that should be answered before any taxonomic change. Some of these may include:
Explain the taxonomic change in detail. Could link to a flag where the explanation is.
If following POWO, has a taxonomic review been done to confirm that the POWO taxonomy is correct?
If not following POWO, has POWO been contacted to confirm why or if they would change? If not, why not?
If endemic to a relatively small region and a taxon would be lumped, do local conservation and taxonomic experts agree with the lump?
How settled is the taxonomy? Are there recent conflicting treatments? Is additional taxonomic work in progress?
Basically, this would show that the proper work has been done before a change is made and how questionable that change may be. More questions could possibly be added to this like whether or not taxon atlases have been checked or updated.
Ideally, voting would not be allowed to start until all requirements have been met. After a change has been voted for, then there should be a notification that goes out to all relevant people that the change is pending and will be implemented in 1 month. Now the notification only goes out after the change has already happened. During that month, further discussion can be had, which could possibly find problems with the taxon change. If fault is found, the taxon change could be canceled or edited to restart the process.
As pointed out over an over again in other threads, the biggest problem is the lack of notification before a taxon change happens. If we didn’t get the notification after that change, it would be even worse. Who knows how many problematic changes would be missed until way too late to fix.
I agree with you, but I do not agree that this interpretation is self-evident. For example, a few people expressed differing opinions on the blog post that I linked above. Some people interpret the accumulation of 5 votes to be the consensus-building mechanism, whereas others interpret the 5 votes as confirmation that consensus has already been reached and that the change is ready to proceed. I see a pretty even split between curators submitting plant taxon changes prior to initiating discussion versus submitting the changes after consensus has formed in a discussion. As far as I’m aware, staff haven’t declared which approach is correct.
I’ve read and participated in the discussion on the blog post linked above, and I am no clearer on this than months ago. If committing a change and waiting for the 5 votes is not the same as consensus building, and if instead the consensus was meant to be reached before asking for 5 votes, then what purpose do the 5 votes serve? If there was already a consensus, why wait for 5 more votes?
If you saw a taxon change page with a bunch of recent comments under it saying things like:
“Hold on! I have no idea how to distinguish these, and I’m well-read on this taxon. What’s going on here???”
“I need help making the atlases complete before we can commit this change!”
“What do people with regional expertise think about this change?” – the taxon change drafter
Would you say that consensus has been meaningfully reached, despite the vote tally reaching 5? Would you commit such a change with incomplete atlasing?
No I wouldn’t. But that just highlights my point. If there’s then a long discussion and eventually a consensus is reached before the change is committed and the 5 vote process initiated, what purpose does the 5 vote threshold serve?
I think a review of the recent premature (i.m.o.) Rubus split referenced above will go a long way toward answering that question.
In the absence of staff confirmation, and in light of recent damage done, I think the prudent and common sense interpretation would be that, by adding the 5 vote requirement, Staff intended the process for vascular plants to become more conservative and stringent, not equally or less so. If they had intended 5 curator votes to replace the more general community input process referenced in the Curator Guide, they would have said so. They were clearly trying to add an extra hurdle, not make it easier, to make taxon changes in vascular plants.
So my 2 cents toward answering this question is:
Patience. Do not begin the community input process by drafting and submitting taxon changes for curator approval. That should be the last step, not the first. (But see my disclaimer at bottom.) It’s already hard enough for the community to find discussions on taxon flags. Discussions on draft taxon changes are even more deeply buried.
Flag the taxon in question for Curation.
Provide/link to as much background information as possible regarding the proposed change.
Tag top observers and identifiers of the taxon (regardless of whether they are also curators) and request their input.
Patience. Wait. At least a week, to give community members a chance to respond.
Assess consensus. In the absence of clear unanimity, how grave are the issues raised in dissent? In a consensus process, when in doubt, the status quo should be maintained until further clarity can be found. Tag additional people when possible, especially if there has been no input at all.
Assess impact. If there has been no input, or there has been at least one strong dissenting voice, how serious and extensive are the potential impacts to the community of going forward anyway? How reversible (or not) are they? Again, when in doubt, consider maintaining the status quo. Think of yourself as the arbiter, not the advocate, for the proposed change.
If going forward, first draft and complete any atlases needed if a split is involved. If help is needed completing the atlases, ask for it on the Taxon Flag, not on the Taxon Change after it has already been drafted.
Finally, draft the needed taxon change(s), providing clear reference to the Taxon Flag(s) where the discussion occurred, and submit them for Curator votes.
Note that changes now automatically display a list of “Curators with most IDs of the input taxon.” In the discussion section of the taxon change, @ tag these curators, both as potential voters, and for any further input if they weren’t included in the Flag discussion. Point them to the Flag and note any significant issues remaining there. Confirm that atlases (if applicable) are believed to be complete.
Curators should only vote to approve the taxon change if they agree that the community input process (including time allowed) was sufficient to the purpose, that the change is justifiable based on the total input provided, and that the changes are drafted correctly and completely and will be committed in appropriate sequence (when that matters).
All this said, I also agree that there are some “no-brainer” changes (orphaned autonyms or other taxa “left behind” by past changes, spelling corrections that can’t be done directly, etc.) where the above process can and should be much shorter, and can go directly to a draft taxon change without a flag discussion. Curators just have to use their best judgement and, when in doubt, err on the side of more patience and community input, not less.
I also agree that this process could be greatly facilitated if iNaturalist would develop ways to automatically notify users (when they have opted in) of taxon flags and taxon changes likely to be of interest to them. But for now, the system is what it is and we have to work with the tools at hand.
Thanks for the thoughtful and constructive reply. These two steps are fairly difficult and perhaps not possible to translate into a one-size-fits-all set of guidelines. The suggestion to maintain the status quo in the absence of clear consensus seems like a good rule of thumb worth incorporating in the curator guide, at least in my opinion.
That said, there are some situations where the status quo is already quite problematic. For example, a curator might add or reactivate taxon A, implicitly narrowing the circumscription of taxon B. (For example, Oxalis smalliana was added in 2023 as a segregate of O. oregana, but O. oregana was not split until December 2025.) Over the months/years, taxon A might accumulate hundreds or thousands of IDs (4,388 in the Oxalis example), while hundreds or thousands of taxon B s.l. IDs that should have been changed to taxon A via an atlased split remain at RG for taxon B. In cases like these, there are often pro-split voices and pro-lump voices, and people on both sides may have invested significant amounts of time. We end up with three paths forward:
Retain a status quo, accepting a large (and growing) number of disagreeing IDs between taxon A and taxon B that could be resolved via an atlased split, and also accept that the range map for taxon B still represents the broader circumscription no longer supported by the taxonomy on the site.
Merge taxon A back into taxon B until there is consensus for the split. This wastes the time of anyone who has been adding IDs of taxon A to correct old observations of taxon A s.l.
Commit an atlased split of taxon B to “finalize” the adoption of the split taxonomy. This wastes identifiers’ time if the atlases include significant overlap zones. This also has the disadvantage of making the lumpers feel cheated, since no consensus was required to create taxon A, and then the existence of taxon A on the site was used as a rationale for the split.
In this scenario, I feel like maintaining the status quo simply allows the cons of either path forward to grow in magnitude over time, unless one gets very lucky and a paper with definitive evidence for or against the split falls into one’s lap. What is our best practice here? Is there more of an impetus to make a decision ASAP, or should we react to a dissenting opinion (perhaps with a pretty good argument) the same way as in a situation where the status quo is more coherent?
I have seen this play out recently with Phoradendron leucarpum, Arceuthobium campylopodum, Lomatiumfoeniculaceum, Rhododendron neoglandulosum, and almost the entire Pyrrocoma genus, so I don’t think of it as an edge case. Or, if it does represent an edge case, these edge cases eat up a disproportionate enough amount of energy to be worth discussing.
I’d like to say, that the above suggestions for problem-solving does not apply to all taxa, maybe only plants and some animal taxa with worldwide found genera or species.
Please do not introduce any such guidlines for arthropods etc. When you tag people on a flag for insects, you hardly get much constructive response from more than one or two users and thus a proposed change sometimes takes years or does never take place.
Also, any discussion should happen on the flag threads, as many people who can give input do not visit the forum. You have taxon specialists who really know a family or genus, but do never show up on the forum.
Definitely not possible. In the end we will have to rely on the good judgement and experience of our curators, which discussions like this should help to cultivate.
In cases where there is no obvious “best” solution, we probably need to lean more heavily on our reference taxonomy. I don’t usually like “POWO made me do it” reasoning, but in some cases it may be the only justifiable action.
In cases where an older taxon change was only partially implemented, and no other “best” option is apparent, I would also default to cleaning up and completing the previous change, even if it means a delayed split that should have happened long ago.
I think dissenting opinions should always be taken seriously and considered carefully, no matter what the context. We are in this place of curator voting etc. because dissenting opinions were too often ignored or dismissed (or never sought out) in the past.
Understood and agreed. This impetus for this discussion (OP correct me if I’m wrong) came out of the current situation with vascular plants, where 5 curator upvotes are required to implement a taxon change. @holyegg unless you wanted this discussion to have universal scope, you might consider editing the topic title to make the intended taxonomic scope clearer.
Also agree that discussion of specific taxon changes should only happen on taxon flags (or sometimes in the discussion area on draft taxon changes themselves). But for more general discussion of best curation practices, this forum category is really the only option available to us.
I think this example may be key to understanding what went wrong in the split/consensus-building process for R. ursinus. I may be missing something (maybe you already listed what I’m proposing next already), but it would seem to me, in the case that there are relatively sharp geographic borders between taxon A/B (or Oxalis smalliana/oregana in your case above) that atlasing and auto-swapping IDs to match what would be correct for where O. smalliana would swap for IDs of O. oregana where it no longer is considered to occur would make the most sense.
In the areas of overlap, I think the “status quo” approach would make the most sense, and, for those unfortunate identifiers that are in the areas where the two species need to be disentangled due to geographical overlap, they’d need to manually go through and confirm or reject the consensus on individual observations.
Alternatively, defaulting to a higher taxonomic level in areas of overlap also sounds acceptable since especially for older observations, turning over consensus may take many years or be in-process indefinitely, depending on how active identifiers are in the area to look back to the earliest observations.
I think in any case, it’s hard to imagine where it’d make sense to revert thousands of observations without clear guidelines on how each species can be identified, and at least some geographic boundaries on where each would be expected.
How is ‘atlasing’ not a requirement for large taxon swaps in the first place? I’ve also seen many cases where swaps or splits seemed to go awry in some major way, and I think it boils down to curators not “doing their homework” or respecting regional identifiers/authorities (e.g. Jepson/FNA over POWO in CA) in many cases.
The relatively easy swaps can create huge headaches and wasted effort that would be averted through simple requirements of atlasing and at least some system of tagging and checking in with the top observers and identifiers (who’ve actually seen the species and all the variations in person) to “kick the tires” on the proposed action for at least a month, as was proposed in the taxon swap thread.
I’ve seen POWO be outright wrong by the way, it shouldn’t be held up as a golden standard that cannot be questioned, which is I think a deeper issue with the curation process beyond lack of consensus-building and blind vote problems.
To be clear, atlases are only used by the system when there is a split. Swaps (1:1 name changes) do not need or use atlases.
Also, I am pretty sure that if someone attempts to commit a split with no or missing atlases, the system throws up a warning message. But if atlases already exist, the system has no way of knowing whether they are accurate or complete. That is wholly on the curator(s) involved. In the Rubus case the split should never have been submitted for voting while the atlases were still in need of revision.
Very much agree with this, but I would note that I have seen some curators drafting a taxonomic change prior to asking for community input because they want it to be very clear what their proposed change is. I think this is actually a good idea because it can reduce confusion about what the proposed change would actually look like in the iNat system (somewhat like laying out a specific hypothesis in a paper before testing it). That said, if curators do this, they would clearly need to create any relevant atlases as well so that users can evaluate those at the same time, as doesn’t seem to have happened in the case study being discussed here, e.g.:
Of course, one potential issue with drafting the change prior to discussion is that someone could then easily go in and submit it very quickly with little notice/without much or any consensus (potentially not even the person who drafted it).
I do think that the voting system introduces some confusion about what consensus means for taxonomic changes on iNat because voting is a common way of measuring consensus. There’s reasonable space for confusion about whether 5 votes is a measure of consensus or not, and the guidelines don’t explicitly lay this out as others have noted. My own position would be that consensus (which doesn’t mean unanimity!) should be considered pretty much independently apart from voting. But clearly there will need to be consensus among at least the voters for the change to be approved, so the distinction is going to be a little fuzzy.
To clarify, atlasing was not the issue in any of the examples I listed—it was the complete lack of a taxon split when the addition of a new species implicitly narrowed the circumscription of another species already widely used on the site.
Atlasing was also discussed in the R. ursinus split, albeit very briefly and off-site on the unofficial iNaturalist Discord. I asked whether the atlases could be refined from state-level data to county-level data, and the drafter of the split said it was not possible. There were atlases in place at the time of the split, which is why R. ursinus in Washington seems not to have been disturbed, but clearly the atlases were wrong or at least very imprecise. That seems like an unrelated issue to the main point of this thread, though.
The “alternative” that you mention in the second paragraph here is how atlased swaps function. The first suggestion, if I understand you correctly, would require removing areas of overlap from the new taxon’s atlas. But, identifiers sometimes refer to atlases to understand the ranges of the output taxa, so I personally would not remove a county from an atlas if I knew that the taxon exists in that county.
It’s worth noting that the curator guide places a lot of the “reidentification” burden on the curators involved in designing a split:
If it’s not already obvious, splits are the most disruptive kind of change. In the worst case scenario where some or all of the outputs are not atlased, observations of the input are going to get associated with some higher-level taxon and will need more IDs to become associated with one of the new outputs. This is probably going to be quite common, so if you split a taxon, take responsibility for your change, review affected observations, and add new identifications where you can (there should be a banner on the taxon change page after you commit linking to a tool that will make this a bit easier).
(I assume this doesn’t just apply when “some or all of the outputs are not atlased” but also when/where the atlases overlap.)
One helpful step in deciding whether to move forward with a split, in my opinion, is to have someone review observations in overlap zones to get a sense for what proportion of observations will be identifiable to species level post-split and to assess the utility of keys. If identifiers aren’t given tools to work with to distinguish the output taxa, they have a right to be frustrated. If the identification tools require analyzing subtle features not easily seen in 95% of observations, then we can likewise expect them not to favor the split. If someone does a little “recon” and finds that the output taxa can usually be distinguished from iNat photos, then there might be some grumbling, but it’s likelier that people will get on board after a short adjustment period.
Taraxacum officinale might be like the final boss of this question as it’s technically taxonomically untenable (not in POWO and not compatible with other species that iNat accepts) but essentially every iNatter in North America and many other parts of the world have a connection with the species concept. There have been lengthy flag and forum discussions about it (which I think are good and necessary so people are at least aware of the issue) but there’s still nowhere near consensus. I can only imagine any kind of recognizable consensus forming if North American Taraxacum taxonomy is significantly updated, which won’t happen for years if not decades.
It’s hard to imagine a more impactful potential change than that but I think it would be beneficial for similar levels of outreach/engagement to occur for some larger taxon changes that have large scale and potential for controversy (e.g. the Ponderosa Pine split last year, or the annual bird updates).