18 October 2021. Languages | Amazon

When a language dies, so does our knowledge of the medicinal qualities of plants; Surprise! Amazon privileges its own products in its product rankings.

Welcome to Just Two Things, which I try to publish daily, five days a week. (For the next few weeks this might be four days a week while I do a course: we’ll see how it goes). Some links may also appear on my blog from time to time. Links to the main articles are in cross-heads as well as the story.

#1: When a language dies, so does our knowledge of the medicinal qualities of plants

There are about 7,000 languages in the world, and more than 40% of them are endangered. In a recent study, researchers at the University of Zurich have concluded that every time a language dies unique knowledge about the medicinal and herbal qualities of plants also dies.

(The full paper is here).

(The language of the Yawalapiti people is currently spoken by only three people. Jean Marconi/CC BY-NC-SA 2.0, via Mongabay)

This sounds as if it might be a marginal problem, but many mass-market medicines derive from plants.

Touching on the details of the research, briefly, it’s fair to say that it was extensive:

The study’s scientists analyzed 3,597 vegetal species with 12,495 medicinal uses and linked this data with 236 Indigenous languages from three biologically and culturally diverse regions—the northwestern Amazon, New Guinea and North America. From this, they concluded that in these regions, 75% of the medicinal uses for medicinal plants are known in only one language.

And the issue is more serious than that, because the languages that were most at risk of extinction were also those which were more likely to hold unique knowledge about a plant—a problem that was particularly acute in the Amazon.

(The dots on the maps indicate the distribution of languages that cite medicinal plants. The red bars show the percentage of medicinal knowledge restricted to just one language in North America (A), the northwestern Amazon (B) and New Guinea (C).)

The conclusion from the study is that cultural loss of languages may well have an even greater impact on medicinal knowledge than loss of biodiversity.

With regard to the maintenance of ecosystem services, cultural heritage is as important as the survival of the plants, as has been previously proven in scientific studies.

We already knew, of course, that loss of a language results in the loss of a particular way of seeing and understanding the world. I’d say that this study makes this more concrete by linking the loss of language to particular types of knowledge of the world.

“Linguists consider a language to be endangered when people stop speaking with their children in their native tongue,” says (Luciana Sanchez) Mendes, who holds a post-doctoral degree from Brazil’s Roraima Federal University. In Brazil, the devaluation of Indigenous languages has been in favor of Portuguese and Spanish—which have been dominant since colonial times—as Indigenous parents forego their native tongues in aim(ing) to equip their children for social success.

Of course, killing off indigenous languages and replacing them with a colonial tongue has been a sport of colonists throughout history: as well as the Spanish and Portuguese, the British were particularly well practised in it.

Perhaps belatedly, UNESCO has declared the decade from 2022 to be a decade of action for indigenous languages. Of course, there’s also a question of power here. The people who speak them tend not to be represented in international organisations, as Jordi Bascompte of the University of Zurich observes:

“These are languages that we tend to forget—the languages of poor or unknown people who do not play national roles because they are not sitting on panels, or sitting at the United Nations or places like that. I think we have to make an effort to use that declaration by the United Nations to raise awareness about cultural diversity.

#2: Surprise! Amazon privileges its own products in its product rankings.

I know I’ve mentioned The Markup several times here over the last month, but they’ve been doing the sort of coverage of the digital tech sector that gets journalism a good name.

They set out to investigate whether it was true that Amazon products got privileged in the way they were displayed to customers by Amazon’s algorithms. Businesses selling through Amazon have long believed this to be the case. However, their complaints were, by the way of these things, individually anecdotal, and therefore easy for the company to dismiss.

In fact, this specific question was put to Amazon during a Congressional hearing last year:

“Does Amazon’s algorithm take into account whether a product is a private label sold by Amazon?”

Amazon replied, in a written response, that its algorithm “did not take into account the factors described … when ranking shopping results.”

The Markup reporters investigating this found that:

Amazon routinely ranked its own its own brands and exclusives ahead of better-known brands with higher star ratings and a greater number of reviews. For instance, shoppers searching for “cereal” would see Amazon’s Happy Belly Cinnamon Crunch cereal, with four stars and 1,010 reviews, in the number one spot, ahead of cereals with better and more reviews, including Cap’n Crunch (five stars, 14,069 reviews), Honey Bunches of Oats (five stars, 5,205 reviews), and Honey Nut Cheerios (five stars, 11,702 reviews).

To test why this might be they worked with a machine learning program, random forest, to build a predictive model of how the Amazon algorithm(s) work, running the model over and over again. The conclusions were decisive, as seen in the chart above.:

Being an Amazon brand or exclusive was the most important factor for predicting the top result of an Amazon search. Being an Amazon brand or exclusive was the only factor needed to make the model perform well. And when information about whether a product is an Amazon brand or exclusive was missing, the model performed the worst.

Other factors like how many star ratings and number of reviews each product had were not even close runner-ups in predicting the top search results.

The magazine has also published its methodology. Amazon’s response? I think the kindest description would be on a scale between pedantry and nit picking.

The summary above is taken from the weekly newsletter of The Markup’s Editor-in-Chief, Julia Angwin. The full story is here.

j2t#189

If you are enjoying Just Two Things, please do send it on to a friend or colleague.