Facebook can't buy a break with its newsfeeds. Every time it changes the model, somebody complains. But its latest snafu — turning over the job to an algorithm without expert oversight is not the answer. Posting a fake story just isn't smart. It's not insights-driven; it's head-in-the-sand.
(The provenance of this image is opaque. If you own it, please let me know so I can attribute it properly.)
- Rule #2: Marry algorithms and expertise to continuously improve outcomes. Algorithms are not a secret sauce; they are a model of the real world. If the algorithm says X and the expert says Y, then there's room to improve either the algorithm or human understanding. Innovators like Allstate Insurance (and Memorial Sloan Kettering Cancer Center) accomplish this by putting product experts (or oncologists) and data scientists in a room to continually refine their cognitive assistants. (see endnote 9)
- Create an editorial board. The board has a very tough job: to decide how best to represent Facebook's position on issues. It may need to show its colors in things like politics. Or maybe not. After all, David Brooks & E.J. Dionne spar healthily and respectfully on NPR every Friday afternoon. It's possible.
- Marry algorithms and expertise. The expert trains the algorithm through the choices it makes. That's means they need a new team — an insights team. This "InsightOps team" combines expertise from data scientists, data engineers, and domain experts. It's what Stitch Fix does to make sure the outfit you get is stylish and desireable: combine stylists and algorithms. It's what Earnest does to price student loans: combine loan officers and algorithms. They are insights-driven businesses.
- Be transparent as why a particular article showed up. How can we trust the algorithm or the expert without complete transparency in how the decision was made?
- Designate an ombudsman to watchdog the entire process. Transparency extends to the editorial board. It's not possible to be perfect. But Facebook must at least identify problems and solve them in the harsh light of day.