Big Data, targeted advertising, medical diagnosis systems, etc. use methods of automatically generating decision trees by recursively searching for criteria that can be used to incrementally reduce uncertainty. These methods can span many dimensions without suffering from the "curse of dimensionality."
An application that I have yet to encounter is to use these methods to: a) analyze social problems and their causes; then b) use the resultant decision trees to write legislation that can adapt to many distinct situations - getting away from "one size fits all" overly simplified laws and regulations.; and c) in the execution of the law use the decision trees to decide on: 1) filing charges or not, 2) severity of the penalties, and 3) appropriate remedial follow up.
There are too many cases of overly simple laws that were intended to ensnare major criminal behavior being applied to citizens who are honest and law abiding, but who are subjected to onerous red tape or investigations simply because some element of their activities or appearances innocently triggers certain criteria that makes them subjects worthy of investigation or inquiry.
This is an excellent proposal that will increase efficiency in delivering up to date and contextually appropriate legislations at national, regional and international levels. I understand the value of data. For instance, I was part of a project jointly conducted by the Internal Displacement Monitoring Center and the Norwegian Refugee Council Pan African Office. The project used data collected from different sources in order to map the causes, trends, unique challenges and potential solutions of internal displacement in Africa and publishes an annual report known as the Africa Report on Internal Displacement during my time there in 2016 and 2017.
But at this juncture may I pose a question and ask how important issues like
- The source of information (independent NGOs, businesses, governments) Who collects such data and how can we ensure such data is representative of the situation in the region and guarantee the production of authentic data?
- How do we reconcile this development with the new challenges in the area of data collection and use which are challenges to the right to privacy with less and less restricted governmetn access to private informationa and communications. How do we make sure there is balance? Where do we draw the line?
Dear Bruck,
Thank you for your considered and knowledgeable response. As I understand it, Big Data analysis uses various data that tend to be unbiased, such as where people live, their buying habits, their age, their sex, their internet usage, ..., that is, many, many attributes. Each attribute, by itself, may not have an obvious relationship with the question at hand; but when combined, conditionally with other attributes may contribute to reducing the uncertainty with regard to separating prescribed categories of (in this case) people. Some categories, for example in income tax collection applications, could be "large-scale evaders", "large-scale money launderers", "small-scale cheaters", "insignificantly small accounts", etc.
I would imagine, if there is a question regarding the reliability of certain attributes, the analyst could, accordingly, assign weights to each attribute so that the more reliable attributes would have more influence on the resultant decision trees.
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Writing