Machine Learning (ML), which I blogged about recently, is beginning to get traction and we think that it has the momentum, for reasons cited in that post, to break out and begin to have a material impact on the way organizations operate.
This difference will apply to public institutions too: this will both offer citizens more efficiently delivered government services and raise privacy concerns.
To drastically simplify things, ML enables users to use computers’ ability to analyze massive amounts of data and identify patterns, trends or, in the case of consumers and citizens, behavior. It doesn’t take a lot of imagination to figure out where governments could apply this capability; let’s look at the more positive ones first.
Government agencies collect a lot of data about weather and that data is used to provide middling forecasts, which are especially valuable when applied to storms. Weather systems are notoriously difficult to model due to the massive amounts of data and complex interrelationships involved. If ML were able to make forecasting outcomes more accurate and do so further in advance of storm activity, BNs of dollars in storm damage, as well as thousands of human lives could be saved.
Related to weather, is the need that farmers have for accurate data, increasingly available at a highly localized level, about the factors that affect crop performance, quality and yields. ML could help government agencies develop broader and deeper data sets that researchers would use to develop crops that require fewer pesticides, provide betters flavor and use fertilizers more efficiently, and help farmers manage their crops on an “ultra-local” basis.
As the country becomes more populated and more developed, the pressure on the environment will increase as will the concomitant need to both understand and manage it better. Ecosystems (Like the weather described above) are fiendishly complex systems with billions of actors and interactions. ML could help researchers understand these interactions better and give government agencies (The EPA, California’s Coastal Commission) better tools with which to make decisions.
It is easy to imagine how the Centers For Disease Control and the National Institutes of Health could use ML to mine that vast amount of data collected on diet, health, disease and drugs to help folks improve their health. Meta studies - ones where data is collected about studies to assess the findings about a particular topic - are common today; one could see a time when meta-meta-studies become the norm.
Those are all pretty much positive examples about how public agencies could use ML to enhance their services, reduce spend and increase quality of life for citizens. Now for some examples that raise the issue of privacy and individual liberty, some of which are already taking place.
Vigilant Solutions photographs license plates of cars, geotags them and sells access to its database (2.2 BN pieces of data so far.) Imagine increasing this dataset 100 fold - say, tracking 95% of the cars in the country with photographs every hour for several years, cross referencing it against local demographic data, driver records and consumer data and then applying ML to understand indicators of criminal intent. If police were given access to this analysis and were able to pull it up for anyone they stopped, they would be able to make the presumption of guilt without any actual evidence. And, they would know way more about that individual than most of us would be comfortable with.
Criminal DNA Data
Most states currently collect DNA samples from convicted felons and this data has been essential in freeing wrongly convicted individuals who would have otherwise spent years in jail or gone to the execution chamber (Good!). However, imagine authorities aggregating this DNA data and using ML to look for correlations between DNA patterns and patterns of criminal behavior. Not too many steps away from running DNA tests on folks and predicting their likelihood to break the law. Or classifying folks by background (i.e. race.)
So, how the government uses ML hinges on the good: consistency, quality of life, efficiency; and the more thorny issues of: privacy, fairness and bias.