How to surface interesting observations

General
1

As an active identifier, at times I find myself struggling to find valuable or interesting observations and often having to wade through a sea of low-quality observations (unclear subject, blurry, repetitive, dark lighting, etc.) to find them. It seems that over time, the problem of finding the interesting observations may become more difficult absent the development of tools to accomplish task of surfacing them. This topic is intended to come up with ideas to surface those interesting observations. If any of these ideas appear to have particular merit, I’d be happy to try to develop it into a feature request.

Broadly, I’d say that there are two general ways to find the interesting observations: either “downgrade” certain low-quality observations or “upgrade” the interesting ones. By downgrade and upgrade, I only mean make those observations either easier or harder for the average identifier to find. All observations have value, but so does an identifier’s time. The benefit of highlighting valuable observations seems to clear to me – more people will see them & sooner than they would under a business as usual approach. The case for downgrading low quality observations is that by removing these from the pool, other identifiers won’t have to spend their time looking at them (unless they choose to, for some reason.)

Ideas for to make interesting observations easier to find:
• Make it easier for identifiers to find species that are newly reported or have low numbers of observations (e.g. <X; X could be 10, 20, or 50+) in a region (could be a US state, country, etc.). Many of these will be incorrectly IDed species suggested by the CV and the earlier they are caught the better. Some may be range expansions or new introductions and also deserve attention. I sometimes come across these observations and if they are obviously wrong I fix them, but many linger for a while and can contribute to wrongly IDed observations getting propagated through an entire region. I am imagining some automated process by which these observations could be searched. I think we all have taxa that we keep an eye on, but there are plenty of taxa that slip through the cracks.

Screenshot (44)
Screenshot (44)1920×1080 144 KB

One idea of implementing this idea would be to have an additional filter on the Identify page. Not sure how much server load this would entail.

• “Upvote” individual observations. Sometimes I’ll come across an observation that looks interesting but I don’t know what it is. I would like to be able to flag it in a positive way so that other IDers would see it as well. I could “fav” it or tag individual IDers but that doesn’t always feel right somehow.

Ideas to make low-quality observations less visible:

• “downvote” low quality observations. If I see a photo of just tree bark, a shot of a lawn with no obvious subject, a blurry photo, etc I mark these as reviewed and move on. But that still means that all the other IDers are going to see this observation and have to spend their valuable time on it. I’m looking for ideas on how to relegate these kinds of observations to the bottom of the pile, so to speak.
o More options in the DQA.
 an option for observations with multiple photos covering different species with no overlap (observations that need to be split)
 an option for an observation with no clear focus
 duplicate observations (e.g. same observer posted 5 photos of the same organism within a short time period)

o If the DQA can’t be expanded, then just an option to “downvote” a low-quality observation. We’d have to come up with a list of reasons an observation could be “downvoted.” Observations marked this way would be harder to find. Maybe if an observation accumulated 3-5 of these downvotes it would go to a separate category. Maybe this option should only be available to IDers with a certain number of observations so it doesn’t get abused.

o If an observation has been “marked as reviewed” by a certain number of people, it goes into a separate pile. Not sure what the cutoff should be 20? 50? 500? I would want this to correlate to an observation’s quality and not just the date it was uploaded to iNat though.

2

Projects have been used to help get things identified (like Flora of X place or pre-maverick observations), since identifiers interested in those subsets of observations can easily browse through them.
If I see a nice plant that I can’t identify, I’ll at least fill in the phenology fields so that maybe some future identifier who’d like to browse flowering unknowns can more ealisy find it.

3

I guess the thing is that one person’s low quality observation is another person’s interesting observation. There are lots of “bad” photos of amazing bugs, for example - and “amazing bug” doesn’t always mean flashy and charismatic. To the average observer these may just look like bad photos of average bugs, but to the right person these observations absolutely will have value - which means they’re probably gonna have to do some digging to find them regardless.

And on that note,

I could “fav” it or tag individual IDers but that doesn’t always feel right somehow.

I would actually recommend you do do this actually - if you know an observation would be interesting to a particular person, tag them. this helps cut out some of the digging time and it gets the right eyes on an interesting observation faster than it otherwise might. Some identifiers might dislike this (and if one tells you to stop then obviously respect their wishes) but I am certain that many identifiers are thrilled to see interesting observations in their target area ; I know I am at least.

4

I was using life + life + good as can be - to get obs out of Needs ID pool (frank duplicates, pictures scattered across multiple obs for one individual) but that was forbidden. They stay in Needs ID.

I have a copypasta for - only a few obs, need 60, and I do tag in people. Absolutely!
https://www.inaturalist.org/observations/197217628

I do always glance at ‘how many obs on iNat’ as I ID. Does this taxon need encouraging - or are there hundreds already?

5

I agree that “interesting” is highly subjective.

I would also be concerned that a system for up- and downvoting observations would end up becoming a mechanism for “liking” aesthetically appealing observations and classifying observations that aren’t felt to meet some photographic standard as worthless. I don’t think photography skill or equipment should be a criterion for equal participation on iNat or a determinant of one’s chances of getting one’s observation looked at.

There is also the question: how would you feel if you saw that your observation had been given x number of downvotes and no upvotes?

People mark observations as “reviewed” for all sorts of reasons that may have nothing to do with the quality of the photos or how rare the organism is. It is not a reliable measure of an observation being “poor” or “uninteresting” or unIDable.

The only thing “reviewed” says is that one particular user chose to take it out of the set of observations they want to look at. It may be that the user does not have relevant expertise to contribute anything to that observation; maybe they are using “reviewed” as a way to keep track of observations they are looking at for some other purpose (say, for adding annotations or observation fields, or as part of a project).

6

We can use “No, it can’t be improved” to get rid of really bad photos. Problem is that, like most people probably, I rarely do that. I almost always think, “Maybe someone else can make sense of this one.”

My use of “reviewed” is not at all a verdict on photo quality. For example, if I’m checking and rechecking observations from a particular area, I usually click “reviewed” for all fungi. This gets them out of my way. That’s all “reviewed” means in these cases. Many of the observations are great, but I have no hope of identifying them – I just don’t have the skills. Also, everything I ID gets marked “reviewed” and that doesn’t mean the photo is good, bad, or other.

7

I sympathize with the Needs ID pile being large (my ID/observation ratio is about 100:1, so I am primarily an IDer) and the frustrations of observations that may not be IDable (or are otherwise less interesting/gratifying).

But I would not want a system that allows additional functions for downgrading/upgrading observations to be implemented. I just don’t think that the benefits would outweigh the costs in terms of complexity, potential for abuse, and discouragement of users.

In terms of “upgrading” observations, I think favoriting and tagging users in observations that match their expertise are both intended uses of iNat and effective. I know some users do look through favorited observations, and many IDers (myself included) respond positively to mentions from other people that they know tag them appropriately. Suggestions to use projects are good as well.

For observations with multiple species, just ID to the most specific common taxon, leave a comment asking to split (and leave a comment when this is done), and move on. If seconding someone else’s higher level common ID, I think it’s fair to tick “cannot be improved”.

One thing that I definitely would like is a better way to deal with duplicates. To me these are the most galling observations, as they really aren’t adding anything to iNat in my mind and are a waste of IDer time. I admit I was a fan of just flagging duplicates as such and moving on - very quick and efficient. Now that that is no longer available, I am a good citizen and leave a comment asking to delete with a link to the duplicated observation, hit “Reviewed” and move on. But I would love a clear and efficient way to take these out of Needs ID.

8

see https://forum.inaturalist.org/t/add-feeds-showing-when-new-species-are-added-to-the-database/48405/3

here’s an example that allows me to look for newly identified stuff in my area: https://jumear.github.io/stirfry/iNatAPIv1_identifications_recent_taxa.html?place_id=110679&quality_grade=needs_id,research&rank=species&category=improving,leading&per_page=200

some related discussion: https://forum.inaturalist.org/t/show-number-of-reviewers-for-an-observation/38623

9

I just don’t think that the benefits would outweigh the costs in terms of complexity, potential for abuse, and discouragement of users.

As someone who primarily just observes, I can safely say I’d be discouraged if someone voted that my observations are not ‘‘good’’ or ‘‘interesting’’ enough to be identified or shown to others, or even worse, if multiple people agreed on that.

10

This is one thing that Scott mentioned in his blog post about the Geomodel - eventually using it to surface observations that may be incorrectly identified or represent really interesting finds.

Faving or tagging other users is generally fine, although it obviously depends on the user and the number of times you tag them. You could also make a traditional project for these observations, as adding observations from the Identify page is now possible. An observation of mine was recently added to Undescribed Diptera, for example.

I don’t think we’ll be implementing upvotes or downvotes. Most of the downvoting goals described here can be achieved via the Data Quality Assessment.

11

I do recognize this experience. The only tool I would really like added to fix this is one that would use CV to find observations with no ID or only high-level IDs that look like a particular taxon, and in fact the excellent @jeanphilippeb has already come up with an ingenious project that does exactly that (see https://www.inaturalist.org/journal/jeanphilippeb/73398).

I know your post focused on the addition of new tools to highlight interesting observations, but I tend to address any lack of observations by trying different search permutations. It goes something like this:

  1. Did I run out of worthwhile observations in my target genus/family and region? First make sure I’m not excluding existing RG or captive/cultivated observations. That usually adds a bunch of stuff to fix.
  2. Change “verifiable=true” to “verifiable=any” to find observations missing a little data.
  3. Change “taxon_id=12345” to “ident_taxon_id=12345” to find observations that may be sitting at a higher level, but already got a relevant ID from someone else.
  4. Now move up to family, order, class, phylum and then kingdom. Each time, set an appropriate lower bound for the rank (this is the difference between looking at every observation in your target area with the ID “Dicot” and everything in the whole dicot branch of the tree of life. Of course, there are some diminishing returns here, so I typically switch to use “per_page=100” or something similar so that I can browse more efficiently.
  5. Now do Unknowns in my target region.
  6. I’ve really exhausted things by this point, so my only options are (a) broaden out to another region, (b) move on to another genus/family or (c) take a break and wait for the next bioblitz to refill all those categories.
12

As mentioned by tiwane, there are ways to move things out of the ID queue that can’t be identified. As a frequent identifier of birds, I’ll mark anything that is too blurry or distant to be ID’d as can’t be improved, with an ID at the lower level I’m certain of (often just Perching Birds). It takes a few extra clicks, but slowly I’ll whittle down the queue of tens of thousands of photos of birds that take up roughly 5 pixels in the photo.

I would like a better way of highlighting interesting observations, regardless of the quality of the photography. Sometimes someone has a question or a comment about an interesting behavior that I don’t have the answer to. If I just ID the observation (especially if I’m agreeing with their ID) no one else will ever see it, most likely.

13

Also, I’d like to thank you for the unending task of identifying thousands of not great photos of Hymenoptera! I’m always unsure what to do as an observer with observations that probably can’t be improved, but aren’t at species or genus level. I don’t know enough about insect ID to check the "no it can’t be improved box (unlike birds), but I don’t need my thousands of observations sitting in the review queue if there’s no way they can be ID’d from photos. Personally I prefer someone with that knowledge to check that button even if it means it becomes casual.

14

I’m sorry, I know this is not going to be a popular comment, but I do believe there are such things as “good” and “poor” quality observations. I’m going through a whole lot of old Unknown and Life obs at the moment and it is undeniable that the vast majority, for one reason or another, are very unlikely to ever get very far through the ID process and/or make a particular contribution to the pool of future knowledge. And it is often all too obvious that the observer has invested little time and effort in the observation… the point, shoot, upload and move on strategy. Of course, someone in the future may find precisely those obs to be useful for something unpredictable, but the possibility is extremely remote, so in the here and now, their main function is to make the few really interesting Unknowns or Lifes ever more submerged and difficult to find.
Then every now and then I come across an observation where the observer has clearly invested a great deal in recording something they found interesting, or unusual, maybe also adding additional information or a question, or the observation records interesting behaviour, or is clearly identifiable if it could ever be brought to the attention of the right people, but I don’t have the knowledge needed to tag the right experts.
Would it be so wrong to reward these observations, together with the observers who really invest energy in their observations, by attributing these obs a certain priority to keep them afloat in the Needs ID sea? Would an incentive to quality (and not just quantity) be such a bad thing for iNat in general?

15

Ive created a project for what may be considered the more interesting moths of Botswana that may deserve special attention from other southern African mothologists as being novelties.
https://www.inaturalist.org/projects/interesting-moths-botswana

This separates them from 17000 Botswana moth observations that just usually need an ID confirmed.
https://www.inaturalist.org/projects/unidentified-moths-of-botswana

Cant ‘special’ observations be placed in their own projects.

16

I agree that I am happy to have my observations ticked “As good as can be” by folks with the knowledge to assess this.

On my profile, I have “If any of my observations should be ticked “good as can be”, please feel free to do so! I’m always happy to learn the limits of what can be IDed, and if there are characters I should focus on photographing to allow for more specific IDs on future observations.”

I doubt many people will read that, but maybe posting that explicitly will allow others to feel more comfortable using that option (when appropriate).

17

I don’t disagree.

The difficulty is – how does one quantify this in any sort of meaningful and consistent way?

iNat already has a variety of functions that allow users to engage with observations that they find interesting: favoriting, adding observations to projects, commenting, following users. Some of these tend to favor certain social factors that don’t directly correlate with the quality of the observation itself.

Any additional mechanism for upvoting would need to avoid the biases of existing options, and I don’t see a way to do that. For example, it seems likely that users would try to use upvoting as a way to give priority to their observations or those of their friends. Even if users were to be excluded from upvoting their own observations, this would still favor those who are well-networked and proactive about trying to attract attention to their observations.

Favoriting: this skews towards good photographs of organisms and situations that are striking, aesthetically appealing, or unusual. Not all users use favoriting for this purpose; many IDers use it for observations that they want to refer to later – because it has particularly good photos of some key trait, or a useful discussion of the ID process, or there are unresolved questions and they want to consult additional reference material and return to it later. And some users will use favoriting as a way of following observations they are curious about and want to find out what it is. However, on the whole, the most-favorited observations are those that have some visceral appeal that is not necessarily connected to the scientific value.

Another factor here is that favoriting tends to be a self-reinforcing process: material that has gotten favorites is more likely to get seen and thus attract more favorites. If one sorts favorited observations in descending order, one sees the ones that have already gotten multiple favorites; if one sorts in ascending order, one sees all the observations that nobody has found interesting enough to favorite. But there is not really a straightforward way to find those in the middle – i.e., those that have gotten one favorite but not enough to float to the top.

Following: users who consistently post observations that other users think are interesting tend to attract followers. These followers may not always be specialists who can help with difficult observations, but on the whole, it means that more people are likely to see that user’s observations and some of them will be able to help with an ID or direct the observation to the attention of someone who can. However, following is partly a metric of social capital (how well networked someone is); it does not directly depend on the quality of the observations.

Adding to projects: The existing functionality is actually quite good for helping to direct attention towards quality observations so that they may be seen by users with specific interests. The challenge here is finding a relevant project that has a solid user base amongst all the thousands of projects on iNat.

Comments: I actually think this could potentially be a useful way to enable users to more easily find observations that are notable. However, at present there is no way to sort observations by the number of comments and there is no easy or practical way to search comments for particular content.

18

why not help to ID for others? We need identifiers …

19

I use the leaderboard. But. Cautiously. Leaning more on familiar and friendly identifiers - who in turn tag in new to me identifiers.

@spiphany you can search - your own comments and comments on your obs
https://www.inaturalist.org/comments

20

Unfortunately I just don’t have an interest in it, nor the knowledge or time to do it effectively myself. I appreciate those who do whole heartedly, and it’s not like I haven’t done any ID’s, but I lack the time and energy to do them in any meaningful quantity.