The Google Images Acorus atrocity

There are a number of absurd results returned when doing image searches of plants. Most well known may be “Mistletoe” where there just as many images of English holly as there are of actual mistletoe, and “pine tree” where the vast majority are, in fact, not pines. But I found something just as disturbing, if not more, when I Googled “Acorus flower”. Here’s what I got:


To clarify, Acorus is a grass-like monocot with flowers in fairly inconspicuous compact spikes, while Irises are the ones with the showy petals. As you may notice, most of the results are for Iris, presumably due to the association of the common names “sweet flag” and “yellow flag.” But if you look even more carefully, a stock “botanical” illustration shows something truly horrific:

Yes, it’s an inaccurate drawing of an Iris slapped together with a somewhat decent Acorus. This illustrator should apologize to both of these fine monocotyledonous herbs and the people offended by this drawing.
Indeed, stock photos and illustrations with catch-all keywords seem to be much of a driver for these image results, but this goes a step further by actually mashing together search results into a bizarre Frankenplant.

Has anyone else come across similarly misguided Google Image results, or perhaps even another unholy hybrid atrocities? Any such examples will, no doubt, be both infuriating and amusing.

7 Likes

Well, it’s one of the funniest things in Google pics, the other for me is searhing for latin names and see how it auto-translates English Wiki in Russian.
And it evolves! I don’t know how you can call an Iris something but not what it is.

1 Like

Frankenplant notwithstanding, there are a few problems with your methodology:

  1. When someone uploads a picture of something, Google doesn’t just use a list of keywords they provide to match against your search terms. They also use the title or name of the file and any metadata–including captions/alt-text.

  2. On top of that, if their image search works like their regular search, the results are partially sorted based on the number of links to that result that get published elsewhere on the web. I.e. things that generate traffic generally have a higher ranking.

  3. Your search terms (acorus flower) are set up so that either “acorus” or “flower” may appear as a match. If you put quotes around both words ("acorus flower") only the full, exact string “Acorus flower” would match, and if you put them around each word ("acorus" "flower") you’ll get matches where both words must appear somewhere in the indexed text such as “acorus flower”, “flower of an Acorus sp.”, etc. which is more likely what you want.

Put that all together and I suspect you’re seeing things like “totally_an_iris.jpg” with alt-text “Yellow irises are my favorite flower!” on someone’s well-trafficked blog, with no mention of Acorus, and suddenly that’s a much better-ranking match than some of the actual Acorus photos.

With the frankenflower, I bet something similar happened to the illustrator, and now it’s just a terrible feedback loop…

3 Likes

The Cornell Lab of Ornithology’s All About Birds website has pages for all the birds species in North America, and they include images of similar species to help with identification. This website is very popular (the top two results on a Google search for a lot of bird species are allaboutbirds.org links for me), so unfortunately 2-3 images of other species often show in the top results, and particularly species that are often mistaken for the species in question.

There’s also a general problem of birds being misidentified in blogs and other internet sources so that any image search for a common but difficult to identify species will show a mix of species.

4 Likes

Thanks for the contribution and the tips!

While I will grant you that this may have been how it all started, the vast majority of incorrect (i.e. Iris) results were stock photos with Acorus calamus in the title of the image and often the caption as well (just as you had mentioned), and Acorus and/or Iris in the keywords. Hence my hypothesis that a lot of this confusion is caused by the common usage of the word “flag” in the common names of both Iris and Acorus. I doubt that it’s a coincidence, given that the odds of such distantly related plants sharing that part of their names is so low, and that no other mislabeled image except for Iris showed up. In other words, there were no sunflowers, no hydrangeas, no petunias, etc.

Also, given that the number of incorrect attributions that came up went down dramatically when the word (flower) was either removed as a search term or replaced with (inflorescence), it’s my hunch that the disproportionate association of “flower” with colorful, showy flowers (as in Iris) plays some role. In other words, most people don’t look at grasses or cattails and think that the bit at the top is a flower or group of flowers. This might (directly or indirectly through the aforementioned metadata, descriptions, etc.) tend skew the results toward the mislabeled Iris camp.

Anyway, beyond the very mild irritation with the results, I actually find these mix-ups by stock image vendors quite funny. I just hope the misinformation doesn’t get perpetuated too much.

2 Likes

The first illustration is signed in Russian, where both plants have different names, but who knows how the author was searching for pics.

no other mislabeled image except for Iris showed up. In other words, there were no sunflowers, no hydrangeas, no petunias, etc.

That’s a big oversight in my original response! I hope I wasn’t being too rude, I just know I’ve been surprised by my own searches that suffered from those issues.

Not at all! I myself am not immune to overlooking such details, then giving a lengthy correction which turns out to be unnecessary or wrong. I’m pretty nerdy and I want to be helpful, but I often get carried away. Just thinking about the times I’ve messed up makes me want to bang my head against a wall from embarrassment

1 Like

Might not necessarily be the same author each time. Looked into it and “air obyknovennyy” is actually the Russian common name of the plant.

Google also keeps a record of every search you have ever conducted (I’m not sure how it identifies specific computers - whether it is based on the TCP/IP address that you are logging in from or whether your device routinely sends other identifiable data). There are many reasons for this - such user data is a valuable resource, for instance for targeted advertising, but also determines what results appear in your searches. Think how valuable this is e.g. to a political campaign where specific messages and even news items can be targeted to individuals who might be outraged by them - Cambridge Analytica claimed that they used targeted advertising to swing the result in a number of national elections.

Fisrt one and the other with Tribulus terrestris are of the same author, and yes, it’s ирис жёлтый & аир обыкновенный, hard to confuse, though leaves are similar and may be the other reason for that.