One problem is the simplistic logic of the algorithms used to make recommendations. If the news or book titles match the preferences specified by the reader, they are considered personalised and of higher value. However, there are many purposes for which we read: to be informed; to communicate with others; to imagine an alternative world. We need different approaches for different reading purposes, and algorithms that not only blend users' data but also providers' data, raising the question: who is distributing online content, and why?
In 2017, there has been an increase of politically charged technology, savvy startups such as DECODE and open-source networks such as The Things Network, where users can represent and share their data. Proponents of these decentralised systems argue that, currently, corporations and governments have access to unprecedented amounts of aggregated data that give them knowledge and the ability to act. Instead, decentralised data-management systems are designed as blockchain networks that can be accessed and managed by anyone and everyone.
Among them is The Hub of All Things (HAT), a set of data-management tools that offer a less quixotic solution to the socioeconomic inequalities of data monopolisation. Devised by academics and backed up by legislation - notably the EU's General Data Protection Regulation, which comes into effect in May 2018, and the EU's updated ePrivacy directive - HAT's personal-data accounts are able to store sophisticated information and hand the personalisation capability back to readers. When combined with personalised adaptive reading, they will be able to deliver new reading-recommendation systems.
These new systems will consider the proportions, sequences and combinations of content relevant for individual profiles stored in a HAT personal data account. For example, a six-year-old girl's book recommendations will be based on her age, gender and literacy scores, as well as the more standard options such as past reading history and language.
In learning-theory terms, the 2018 reading algorithms will move from hierarchical individual models (think of a typical classroom setting) to distributed individual models such as Massive Open Online Courses. These models will be controlled by individuals, not governments or private corporations. However, unless they can offer a lot of highly specialised content, they will have a high attrition rate. This is because distributed individual models work best for self-motivated, confident and capable individuals. Those who struggle to use the internet will struggle to personalise it. We will therefore need distributed collective models of reading algorithms. In these models, readers will feed off each other's knowledge and support each other to deal with fully editable recommendations based on their reading history.
This article was originally published by WIRED UK
