Algorithms’ maddening mystery
SOME family members were unsympathetic about my aggravation over arrogant algorithms.
It’s your own fault for clicking on crap was the gist of their succinct message.
Let’s be clear. I hadn’t fallen down some rabbit hole which started with cute cat videos and ended with conspiracy theory supremos QAnon.
No, it was my longstanding subscription to a mainstream news organisation causing the grief.
I noticed when I scrolled down their website there was a ‘‘Hello Elspeth’’ and some articles recommended for me based on my reading history.
Maybe this had always been there and I had been too stupid to see it. Too busy clicking on crap, possibly.
The faux cheery greeting was irritating enough, but then I became incensed they were trying to tell me what I might want to read.
I am not entirely stupid. I know about clickbait and that algorithms on the internet are used to attempt to sell us stuff. (Still, having ads pop up for
The Trauma Cleaner months after I had bought the book online made me wonder if someone somewhere was trying to drive me to commit a crime which might make me require the services of such a cleaner. Spooky.)
But when a news site suggests it has worked out what I want to read, I feel it’s like a zealous librarian pushing me in the direction of some Dewey decimal number before I have had a chance to tell them what I am looking for.
What have I clicked on that tells them I am interested in ongoing coverage of The Masked
(expletive deleted) Singer? Is it Simon Bridges’ fault? Why has their clever clogs algorithm not noticed I have read one of the suggested stories already, and on their website? I read it again because I wondered if it was different. It wasn’t. Who knows what that did to the algorithm’s calculations?
Because I can be as irritating as a presumptuous algorithm, I tell the news organisation I am a grownup and do not need their help choosing what I want to read on their website. I tell them I would be grateful if they could eliminate this feature which I consider is designed for their benefit not mine.
They tell me they have no facility to turn off this option at this time. They do appreciate my feedback because their teams are always looking for ways to improve their customer experience. I tell them they could improve my experience by stopping ‘‘Hello Elspeth’’.
The use of unseen algorithms is not confined to private companies, of course. The Government claimed a world first when it launched the Algorithm Charter for Aotearoa in 2020 for public agencies using algorithms. The idea is that it will help them to strike the right balance between privacy and transparency, prevent unintended bias and reflect the principles of the Treaty of Waitangi. Its focus is on the uses of algorithms with a high or critical risk of unintended harms for New Zealanders.
That sounds important to me. What is surprising, then, is that now the charter has been reviewed for the first time, that review appeared without fanfare on the Government’s data.govt.nz website some weeks ago. When I queried the silence, I was told further promotion by Stats NZ and the minister is likely in the coming weeks, a vagueness I have come to expect from those dealing with my inquiries over this.
The review shows some agencies have struggled to find the expertise they need to measure bias, which the reviewers describe as a key feature of algorithmic oversight.
Some agencies had conflicts about transparency because that could indirectly affect the operation and effectiveness of algorithms, such as those that supported identifying criminal activity.
The subject matter experts interviewed for the review pointed out that at the moment there is little opportunity for New Zealanders to get individual recourse on decisions made about them as the result of an algorithm. They said in the United Kingdom judicial review had been used to challenge decisions made based on a facial recognition algorithm. Here, in the case of a privacy complaint an individual can go to the privacy commissioner, but there is not anything like that for algorithms.
It was good to see those who participated in the review recognised the need for more proactive engagement with the public to build public awareness and trust. Converting documentation into plain English was, however, seen as challenging.
There is clearly still much to be done.
As for my algorithm issue, the naughty child in me wants to click on only the silliest, most trivial stories I can find on the news website to see what that will do to the ‘‘Hello Elspeth’’ recommendations. Maybe that was the plan.