These terms are actively framing adjustment to the algorithm in an unnecessarily negative / inaccurate light. There is no “force” in the suggested change, any more than the existing algorithm forces one thing over another. For me I would say the existing algorithm “forces” a more illogical outcome than necessary. I could also argue it “erases” the finer ID of taxon specialists as it stands at present.
I still hold by the funneling algorithm as making sense in the example you said in the post above, as well as in the other thread where @matthewvosper walks through it.
As I understand it @tchakamaura also states this as optimal…
I think the three of us are in agreement as I understand (?)
Regarding other versions…I confess to being a bit lost in the quagmire of dialogue here …(and short on time this evening to read entire thread). Ultimately, the specifics of implementation are debatable, but you didn’t respond to the comment I tagged you in :
?
Fundamentally, this is the core question for me.
I don’t really understand resistance to this.
Ignoring precise changes or implementations… this fundamentally doesn’t make sense as it stands. Surely you agree ?
If 1000 people say X is a bird. 1 person says X is a fly. 1 person says X is a sparrow.
Surely you don’t believe the fly ID should be weighted against the sparrow ID here?
Surely at some point there has to be a cut-off for you where it goes sooo far beyond the realms of logic it’s not worth entertaining?
as in the previous thread, you’re focusing on the benefit for your particular case and completely discounting the negative effects for other cases. my answer for you is the same here as it was in the last thread:
if i recall correctly, in that other thread, you outright dismissed some scenarios i posed to you to demonstrate unintended side effects. i believe you characterized those scenarios as being unrealistic…
here’s my courtesy to you to go through a version of your similarly unrealistic 1000 birds scenario:
if you have 1 Red-Winged Blackbird, 1000 New World Sparrows, and 1 Savannah Sparrow, the best approach should give you a community ID of New World Sparrow because it should be evaluating 1001 IDs of New World Sparrow vs 1 Red-winged Blackbird. It would be wrong to discount the Blackbird ID and get to New World Sparrow based on 1001 IDs of New World Sparrow with no disagreements.
if you have 1 Red-Winged Blackbird, 2 New World Sparrow, and 1 Savannah Sparrow, the best approach should give you a community ID of New World Sparrow because it should be evaluating 3 New World Sparrow vs 1 Red-winged Blackbird. It would be wrong to discount the Blackbird ID and get to New World Sparrow based on 3 IDs of New World Sparrow with no disagreements.
if you have 1 Red-winged Blackbird, 1 New World Sparrow, and 2 Savannah Sparrow, the best approach should give you a community ID of New World Sparrow because it should be evaluating 3 New World Sparrow vs 1 Red-winged Blackbird. It would be wrong to make this a research grade Savannah Sparrow based on 2 Savannah Sparrows with no disagreements.
note that in this last scenario, that other approach that i thought tchakamaura was describing would give you a community ID of New World Sparrow, and (I think) a non-research-grade observation ID of Savannah Sparrow (not research grade because the community ID is not Savanah Sparrow). to me, it’s not an improvement upon the existing algorithm, but it’s less objectionable than ending up with a research-grade Savannah Sparrow observation ID here.
Imagine all identifications given to an observation. Make a tree out of them, and navigate the tree by picking the branch with most support until no branch is supported by greater than [insert cutoff]. Stop here for the CID. For the Observation ID proceed if the branches don’t actually bifurcate after that point (because there are no alternative/disagreements), that is, continue until there is a split. Return the group with all the options at the split. Thus the Observation id will always be nested within the community ID.
Current classification systems are hierarchical and this is a/the way to sensibly navigate them.
What should not ever happen is just looking at tips, and holding off on judgement if they don’t all agree.
The whole tree stays associated with the observation, and these decisions are re-made whenever a new identification is added. Nobody is excluded manually by any force of will. The transition from a case of 2:3 to a case of 1000:1 happens seamlessly, without any exceptions or special cases. And it is reversible at any time. All the more so because people are required to add fewer id’s in some cases, particularly if they want a low-level taxon to show up, so there is less to overturn. It also doesn’t require babysitting, i.e. posting an id too specific for your knowledge but agreeing with someone else, waiting until it is id’d by someone else at a high level, then retracting your id, which is a way it has been explicitly said that people deal with this. Since there are fewer id’s total, the amount needed to change the location of the CID does not grow as fast with each change.
Even without the Diptera ID
the only consensus is 4 at Not A Fly superfamily.
But without the Diptera it would show a display ID (up top) at Genus - and still with the discreet CID on the right column determinedly staying at superfamily consensus.
If you @mention successfully for a third ID at family, you could move the CID from superfamily to family.
If research grade requires that the CID matches the Obseration ID, then this would not be research grade. I thought Research grade happened whenever two ID’s agree uncontested.
Then we have to agree to disagree. :)
I think in these instances one should indeed discount the blackbird ID.
1000 people agree it’s not a blackbird!
I don’t necessarily disagree with your other examples though.
It’s not about the specifics of whether it becomes RG to me
That’s a separate debate to my mind.
The question here isn’t about the exact implementation.
It’s about acknowledging there is a cut-off point where the algorithm abandons common sense.
I am also ok with it impacting CID but not OID.
Without reading it through, that’s not my memory - so again I feel there might be some slightly unhelpful hyperbole/mischaracterisation in play here. But even the relatively unrealistic cabbage human one, Matthew walked through and I agree with his summary of how it would turn out.
Your example.
I found the next taxon specialist - checked that they cover that location, that they are still active on iNat. @mention = second taxon specialist ID.
Which I supported.
It is now at Genus as you wish.
If you use the iNat Enhancement Extension you can see
taxon specialist one 2189 IDs
taxon specialist two 510 IDs
and my one to convince the CID to work with us please!?
I will make this system work for me, and find a way to work within the constraints of iNat working as intended. (PS fair warning ‘supporting taxon specialist’ is not within guidelines, but it works)
PPS what is more interesting with your example. If you open up the CID algorithm it actually shows 3 Athalia against 1 Diptera (human logic, thanks)
Back to the observation page it shows - Cumulative IDs 7 of 9 for Athalia (what is that about??) If 8 is the Diptera, what is 9?
i’m sorry, but you’re all over the place. you can’t have things go one way some of the time and another way some of the time.
the funneling approach that you say you like will give you a community id of Savannah Sparrow and an observation id of Savannah Sparrow. you would end up with a research grade observation.
you can’t say that you like the funneling algorithm and then disagree with the results it will produce.
you also can’t have things work one way some of the time and another way the rest of the time. the the system can’t read minds to figure out which approach gives the result that sbushes wants in each case. you have to have a consistent approach. if you don’t like the side effects either, i don’t see why you keep trying to push for the approach that creates these side effects.
This is my algorithm ,and should be counted as my suggestion:
Let’s say RG occurs when the following criteria are met: CID and Observation ID match, and the identification is at species level or is voted ‘good as it can be’, etc. If it is, then it will never be the case that RG happens unless the CID itself got pushed all the way to species, which means there really was either an overwhelming vote, even at the species level, because there needs to be > 2/3 votes for the CID to move forward. So I don’t think you need to worry about things getting to RG spuriously.
There seem to be a number of misunderstandings (for the whole thread):
This post isn’t explicitly about ‘research grade’. Deciding that is another issue. People keep wondering ‘when will this happen, when will it reach RG’, but that is a decision to be made, like 'it will do so when CID and Obs ID match, and also they are below some level (e.g. species).
There is no conflict in moving the observation ID one or many steps ahead wrt to having a community id where: ‘the id should reflect the highest point we can all agree on’. I have seen this offered as a counter to the former, but they are not in conflict, because the id’s are separate.
With a community ID at a high level - whatever the observation ID - we are already saying ‘we all agree this much but no farther’.
The Observation ID is not a measure of consensus. It explicitly moves forward without consensus, all the time, by design.
How is RG defined in the first place?
My algorithm is about CID and Observation ID. This algorithm makes sure that the two ID’s behave sensibly wrt all the identifications offered, and also that one is always nested in the other.
This proposed solution (e.g. to the fly problem) tries to be true-er to what it means to identify something as ‘bird’, which is that it also means ‘not plant’.
i’m sorry, but you’re all over the place. you can’t have things go one way some of the time and another way some of the time.
I find this to be mean. I only weighed in in response to a direct question (by you) about whether things would get research grade. That it seems to be a large concern of yours means I find your framing of this as my waffling a bit disingenuous. I did not want anything with regards to ‘research grade’, I merely misunderstood it.
