Thread by @ginasue

So is today a good day to talk about the gap between how AI & algorithmic systems work and how people SAY they work?

A thread about our new report 1/12

Last week we launched AI@Work, a report on the past year in the structual challenges of AI implementation. oii.ox.ac.uk/blog/ai-work-o… 2/12

What we find is a gap between the social and technological infrastructures to make AI systems work fairly, safely and effectively. 3/12

Time and again, we see organisations making the same mistakes in the integration of AI into their decision-making: Over-reliance on the tech, poor integration into the larger data ecosystems, and lack of transparency about how decisions are made. Sound familiar? 4/12

Our study really honed in on the impact that this AI implementation gap is having on workers and on workplaces: AI is creating new problems and extra work, obscuring the work of human labourers, and creating long global supply chains of data and work to build faux automation. 5/

For all the AI 'magic' though, the one takeaway that rings loud today is that AI systems often make binary choices in complex decision environments. Again, sound familiar? 6/12

The report is organised around analysis of published cases. That Amazon HR hiring tool that discriminated against women? Sure bad biased data, right? The system was designed to cut off all the top resumes, reducing a complex hiring choice into a set of binary cut points. 7/12

That brings us to today in the UK with the disastrous implementation of a machine learning system to adjust A-level results.
bbc.co.uk/news/education…

8/12

Step one for AI accountability is allowing people the to push back on the results. That absolutely should also include workers and workplaces. 9/12

In our report we recommend fixes to the problems of transparence, over and under reliance and to integration into social systems. This includes closing that gap between talking about AI as revolutionary and accepting the very real limitations of today's systems. 10/12

The full report is here. Let us know what you think. cdn.futuresays.org/content/upload… 11/12

Thanks to @futuresays_ @biancawylie @juliapowles for supporting this work and to my brilliant co-authors @maggieselfie and @NayanaPrakash1. Thanks to co-conspirators on the Future Says site -- please join us to help make a better future! futuresays.org/empower-worker… fin.