Trying to optimize identification coverage

Trying to optimize identification coverage: there are three main approaches to identifying, each with its own advantages. Ideally, we want to achieve a balance of approaches to cover the data best. Understanding the numbers of reviewers using each approach may provide useful calibration data.

Descending order (newest first) provides encouragement to new users by getting their observations reviewed quickly; and it is also encouraging to longtime users for the same reason. The disadvantage is that it may not reach observations more than a few days or weeks old.

Ascending order (oldest first) helps to clear observations that have languished a long time, and helps to fill in a longer timeline of data. One of the disadvantages is that many of these observations are more difficult to identify, so it takes more time to provide the same number of identifications.

Random order helps to fill in the gaps; if a given taxon has hundreds of pages of observations, the ones in the middle may not be reached by reviewers who go either ascending or descending.

Of course, many of us use different approaches at different times; this poll is to find out the numbers of reviewers who most often use each approach. So if you spend more time on one of these approaches than another, please vote for the one on which you spend the most time.

  • Descending order (Newest first)
  • Ascending order (Oldest first)
  • Random order

0 voters

1 Like

Order is not everything, do you filter for date observed, uploaded or updated?

3 Likes

I sort by Date Updated for newest, Date Observed for oldest

1 Like

I use descending order most often. I use ascending order specifically for identifying unknowns for new users within the last 7 days outside my normal geographic range (which I have been working on periodically in the winter months). Some reasons:

  • I want to catch the new users with some amount of helpful explanation before the clock expires (I do realize I could manually adjust the timeframe).
  • If someone more local is monitoring the unknowns, I would rather them have a chance at making the ID. They are likely to a better job and retain knowledge from subsequent IDs.
  • I try to avoid IDing unknowns too quickly (within an hour) after they have been posted. Maybe they will fix it themselves or are even still working on it.
3 Likes

I am pretty thorough with my IDs, so I only ID groups/regions I know I can complete and keep up with new observations. I usually start with ascending, but when I’m just keeping up all the observations are new anyway.

7 Likes

I had been doing newest first, but only recently switched to oldest first as I’ve been staying only a few days behind the queue of Hawaiian observations. That way, other identifiers can hopefully grab some of the easy stuff and leave the harder things for me.

I sort by random when I’m identifying unknowns.

2 Likes

Newest first but I try to tweak my URLs by using filters, so I can hope to get all the way back to the oldest. The embattled old, the shiny new, and the squeezed middle - all need IDs, preferably soonest.

1 Like

Lately I’ve been doing random but with an observed-on date at like the beginning of the month, to catch any recent stragglers.

4 Likes

I usually ID within projects and start with newest first, then sample randomly if I’m up to date on the latest observations. I only just figured out from this discussion where the option to sort by Ascending date is, so I haven’t been doing that, although for small sets of observations sometimes I will start on the last page.

Aside from being the default setting, Descending date also gives me a better chance of finding out what the host plant is or if the observer has other photos that show more information about the organism. Although I have come across a few people with older observations who turned out to have excellent notes!

4 Likes

I typically do newest first, but I am doing 100% reviews by state so I do get to the older stuff. And like you said, a lot of the older stuff is harder to ID (photos taken with older tech equipment and we know more collectively now about fieldmarks to get than we did).

However, setting it to random periodically does break it up a bit.

6 Likes

I don’t really have a preference, but since in my area a lot of observations go without identifications and older ones tend to be overlooked I’ve voted for Ascending Order.

New observations are, obviously, well and good, but it’s important to get those older ones identified.

4 Likes

I’ve tried ~all the ways, but I think my groove is when I’m in a back page getting toward the 10K display limit- an approximation of “toward the middle” for the number of >20K obs in a typical pile. You can go either forward of backward from there (desc vs asc), looking for whatever you’re interested in, and not get in anybody’s way. An easy accessibility slice of 20K is enough for a day’s id’ing. ;)

Doing random is a good way of not getting in anybody’s way, but you lose some context (school groups ftw!) and it gets laggy for me sometimes.

5 Likes

Descending. However, I’ll often check out a particular date in the past. A year ago? Other? Lately I’ve been going through July 2020, for no particular reason except that northern hemisphere summer has more diversity than winter.

2 Likes

Random is definitely the least preferred way, you can’t see a context, will certainly miss duplicates, etc.

4 Likes

I like to be able to see page count, so random bothers me in that regard

4 Likes

For me, it varies. When I’m working on IDs for my state, I tend to go back several pages. (Especially in my state’s odonata group. There are a bunch of observations there that have been languishing in ‘needs ID’ for two years and more.) I may not be the most expert out there (far from it, I’m sure), but even if all I can do is knock out some of the low-hanging fruit, it helps.

If I’m just exploring to see what other people have seen lately, I tend to go newest first; even then, though, I’ll surf through several pages to make IDs where I can.

3 Likes

Probably only applicable to non-animal life, but I prefer (newest first) because, aside from encouragement for new users, if something is missing that would help with ID:

  • I can ask if the observer if they noticed X
  • I can ask if the observer can add more photos for to see X
  • Others can visit the location and take more/better observations.

All of this surprisingly happened with an observation of Fumaria parvifolia, a non-native in the area where I observed it:

  1. Initial visit
  2. Revisit based on feedback
  3. Someone’s better observation
3 Likes

It is an interesting family. Apart from Fumaria we have our own Cysticapnos

My gut feeling is - that current distribution of 50% new, 30% old, and random covering the gap - should be optimising identification. What would be a good / ideal distribution @tiwane ?

The oldest obs for the Western Cape have often been trapped in the gap transferring from iSpot to iNat. The ID is there, but hidden in the comments - if I make the formal ID and @mention one or two = Research Grade. A huge plus is the MANY species that have been added to CV month by month, making it easier to move them on a taxon level or more now.

2 Likes

Interesting question. I think I use descending and ascending order, but rarely random. Essentially, I change things up in whatever way I can to help keep me interested in IDing longer.

4 Likes

For the category I really have the most fun IDing (plants at a high level) I’ve gone random IDs as my “default” I really like sorting into Family/Genus bins from Kingdom-Class without location preference. It ends up that I look at more observations than I can ID, sometimes way more, but there’s no other way I see of doing the same group and still find it fun. If I sorted in order I’d have big stretches of flora that I don’t know at all, which is tedious, or I’d have to limit the search significantly which reduces the chances of seeing cool oddballs which I always find exciting.

The amount of IDing I realistically do means that I’ll never finish the 200,000+ needs IDs in that search so being comprehensive is not really one of my goals, just moving IDs forward.

I’m really in the “casual IDs for fun” category compared to most folks in this thread though.

4 Likes