Don’t Build Models Requiring Non-existent Data

by | Jun 3, 2021 | ICT4D |

models and non-existent data

When you buy a lego set, you expect all of the pieces to be in the box to make whatever the kit is for. This is a reasonable expectation. If you decide on a making a certain meal for dinner, it is reasonable to think that you either have the ingredients already or can get them. So when a predictive analytical tool gets built which requires data that is either unavailable or doesn’t exist, it is odd and sad.

I suppose one answer is ‘we need to go get the data that doesn’t exist.’ But perhaps a better idea would have been to build a model that works with the data we have.

Perhaps this is one more reason to design and build things in the context in which they need to be used. Building software in contexts where fibre broadband is considered slow is unlikely to work well in communities relying on dial up. And the solution is not ‘well, they should have fibre.’

The choice is up to us.

Photo by Suad Kamardeen

0 Comments

Submit a Comment

Your email address will not be published.