This week we were reminded that many of the problems we’re seeing emerge with tech are just new iterations of problems we’ve always faced.

Alexis Magrigal in the Atlantic looked a series of killings widely reported as “caused by WhatsApp”.

“This year has been presented as an epidemic of violence, aided and abetted, even caused, by WhatsApp. The narrative slotted neatly into the broader discussion of Big Tech’s failures, the corrosiveness of social media, and the crises of misinformation across the world. After all, WhatsApp usage has exploded in India over the past few years, across city and country, rich and poor. Two hundred million Indians now use WhatsApp. Communal violence has been on the rise, going from 751 incidents resulting in 97 deaths in 2015 to 822 incidents and 111 deaths in 2017”

He makes a compelling case that many are just bad humans doing what they’ve always done, just now on WhatsApp.

Another paper published in Science points to the fact that “fake news” constitutes just 0.14% of an average American’s daily media diet.

Closer to home we had several reminders that disinformation and sensationalism is a wider societal problem, with 2FM looking for a debate on the efficacy of masks, and RedFM broadcasting advice that Vitamin C can cure Covid.

I don’t highlight these to encourage us to treat tech related (or tech enhanced) problems any less seriously or urgently, but just to remind that a narrow focus on just the tech involved can blind us to the deeper root cause, making our diagnoses and ultimately our solutions less effective.

📰 News

Are Twitter’s Images Biased? People can upload images of any shape or size to Twitter, but Twitter crops a small section of the image to use as a preview. In 2018 they introduced “smart auto-cropping“, a machine learning algorithm that tries to use the most interesting part of the image in the preview. Several users noticed this week that, in images with black people and white people, the algorithm seems to always focus on the white person. EG 1. EG 2.

Twitter responded fairly quickly, saying that they did test for bias when they launched, but they’ll investigate again. Another person did a test with a random sample and couldn’t recreate the effect, so it may be a case that the only examples becoming popular are the ones that seem to show bias, rather than the ones that don’t.

The whole incident is a great example for why we need more frameworks for auditing the results of automated processes. Their effect matters far more than their intent. In this case Twitter’s engineers just wanted to know “what’s interesting?” in the image, but some checks for “Is this automating racism?” “Is this automating sexism?” would still have been useful. Link.

Co-working. Apparently there’s no 400 remote working spaces in Ireland, mostly in rural areas. Rachel Lavin profiled a few of them in the Business Post. Link.

Brexit. The Johnson Government are detailing their post-Brexit plans to pick the future winners in emerging technologies, give them state-aid and create a world-leading British tech sector. Besides the difficulty this will cause in reaching a trade agreement with the EU, and the fact that all of us would love the clairvoyance necessary to pick the next Google, there’s some merit in direct funding of new technologies with public benefit. The US tech sector benefits from a lot of R&D done as publicly funded military funding, so they could achieve similar effects here, if done through the NHS for example. Then again, you probably don’t need to leave the EU to do that. Link.

💡 Interesting Links

Misinformation. An interview with Bill Gates about misinformation online. Over 20 mins he highlights many of the areas where we know the problems, but don’t yet have many concrete solutions. Link [YouTube].

Fixing bias. Algorithms can help us uncover and address systemic bias. Here’s one good example from France, where the allocation of daycare places is in the gift of local politicians, who are fighting against the algorithm that will do this more fairly. Link.

Oldspiracy. Forget 5G, in 1890 the New York Herald reported a theory that Russian Influenza was caused by the newfangled electricity. Link.

The first death sentence over zoom. As if 2020 couldn’t get more grim. Link.

Buying myself back” Model Emily Ratajkowski talks about what it means to own your own image. An excellent read. Link.