iNaturalist’s flaws can leave people with a false impression of the natural world around them,Sarah Hearne,The Skeptic

While walking across the Scottish moors recently, a friend of mine did what many of us now do. He took out his phone to take a photo of the heather that surrounded him, and asked iNaturalist to identify it for him. Imagine his surprise when instead of being surrounded by Scottish heather, Calluna vulgaris, he found he was apparently standing in the middle of miles of tree heath, Erica arborea, native to the Mediterranean and East Africa. Had my friend stumbled across an invasive species that had taken hold of the highlands with no-one noticing?

Artificial intelligence-based apps like iNaturalist, eBird and FloraIncognita have rapidly grown in popularity. Now anyone with a smartphone can be an amateur naturalist, identifying species and recording their locations with a couple of taps of the screen. These apps have been praised for their ability to resolve the mass of “wildlife” we see every day into distinct and knowable species, helping us connect with our environment and grow our appreciation for and understanding of the biodiversity all around us.

But are these apps everything they’re made out to be? My friend’s experience highlights one major limitation – their ability to confidently identify something incorrectly. Were my friend less sceptical or more absent-minded he could easily have accepted the identification and gone on his way feeling concerned that the moors were covered, not in their traditional heather, but in an introduced and clearly invasive species.

It is a frustrating fact that not every species can easily be identified by unaided visual examination alone. This isn’t a problem just for the apps, of course, but it is one that these apps often obscure. A traditional field guide will present images of species along with standardised identification information, often describing key areas to examine in order to differentiate between similar-looking species. More specialist identification guides offer dichotomous keys, which present the user with a series of binary options about characteristics which will lead to the identification of the species.

In contrast, ID apps present you with a selection of suggestions with a range of certainties, and asks you to pick an answer with often little additional information to go on. If you’re working from a key or a field guide an uncertain identification is easily, if frustratingly, resolved with an “I give up!”. But when presented with an artificial intelligence telling you that it’s likely you’ve got a particular genus or species it’s very easy to just say “sure, why not” and accept the identification.

Many invertebrates can only be reliably identified to species through dissection and microscopic examination of genitalia. Likewise, mosses and lichens require microscopic examination, sometimes with chemical staining, for many species to be accurately identified. These requirements make it very hard for those without significant resources to study these groups and they end up languishing, unstudied. These barriers to identification sets up a “reverse bootstrapping process” whereby easy-to-describe groups get more attention and hard-to-describe groups get less attention. Again, these problems aren’t unique to the apps, but they do exacerbate them as people are focus on taxa where good identifications can be guaranteed while those without are ignored and the AI gets much less of a chance to improve its recognition abilities.

A fungi identification app was linked to several hundred poisonings in France last autumn due to misidentification of toxic mushrooms as edible.

The apps have ways of ensuring that misidentifications don’t make it into databases, and any researcher worth their salt would double-check any obviously odd records before including them in their analyses, but for people like my friend who use these apps for their own edification rather than to be a citizen scientist, it is very easy to be left with a complete misunderstanding of the wildlife around them. That misunderstanding is usually harmless, but there are instances where it may not be. I myself have had the poisonous hemlock water dropwort, Oenathe crocata, misidentified by an app as the edible herb parsley, Petroselinum crispum. Fortunately I did not accept this identification, but with the increasing popularity of foraging, others may. A fungi identification app was linked to several hundred poisonings in France last autumn due to misidentification of toxic mushrooms as edible. And a study published last year by Jenna Otter and colleagues that tested the ability of three different identification apps to accurately identify toxic plants found a wide range of accuracies and consistencies, leading the authors to caution against using the apps as a stand-alone tool.

It’s important to remember that species identification is not the only aim of these apps. Connecting people to their environment and each other is a key goal for many. A thought-provoking paper by Soledad Altrudi published last year examined how well iNaturalist achieved these aims. Through a series of interviews the author reveals a complex picture. While some users find the app has enhanced their understanding of their natural world, many others admit to increasingly relying on the app to identify their specimens rather than studying them themselves, and to focusing on getting photos that will garner them attention from other users, rather than on improving their understanding of their environment. Getting a photo and a name is the aim. The app, rather than acting as a portal to learning about the organism, is a barrier.

The paper also raises the question of whether we even need to know the scientific names of organisms to appreciate them. If you are attempting to collect data for one of the numerous recording schemes that occur in the UK and around the world then accurate scientific names are vital. But identification is the beginning of the process of understanding our planet’s biodiversity, not the end. And for most people, it is an entirely optional step in the process.

It doesn’t matter if the bird you watch build a nest and raise chicks to successfully fledge is called Cyanistes caeruleus, blue tit or Bob, it’s still a magical thing to follow. It doesn’t matter if you call it cleavers, goosegrass, sticky willies, Gallium aparine or ‘that sticky plant’, but you’ll probably know it’s great to stick on other people, preferably somewhere they can’t easily reach. Nature is far more than just a collection of names. It is behaviour and ecology and interactions and folklore and history and myth and legend, and we ignore that to our great loss.

The post iNaturalist’s flaws can leave people with a false impression of the natural world around them appeared first on The Skeptic.

AI-based apps like iNaturalist are fun but flawed – they’re no substitute for real interaction and engagement with nature
The post iNaturalist’s flaws can leave people with a false impression of the natural world around them appeared first on The Skeptic.