Issues in Designing News Selection Algorithms

The post-lunch session at ECREA 2014 today starts with panel chair Michael Latzer, introduces the role of algorithms in shaping our reality and guiding our actions. There is now a range of algorithmic selection services which shape our consumption choices; these include search applications, aggregation, observation and surveillance, forecast and prognosis, filtering, recommendation, scoring and reputation, automated content production, and allocation (e.g. computational advertising) applications.

But the first speaker is Sean Munson, whose focus is on news algorithms. Back in 1970, some 50% of US adults watched the nighly news broadcasts; this percentage has dropped off, in favour of regional, ideological, fake, thematically specific, and other niche news services. A majority percentage of users now use news aggregator sites, and many also draw on social media for their news; this may have created the 'Daily Me', but may also lead to the development of disconnected filter bubbles.

Algorithmic designers can build those bubbles, or break them. Predictive news aggregators may filter based on existing news preferences, locking users into their past behavioural patterns; social filtering may lead to encounters with material of interest to socially homologous groups only. This may end up with fewer opportunities to learn, less common ground for decision making, fragmentation, reduced empathy and understanding, and lower engagement.

On the other hand, such barriers may be broken by embracing different selection strategies. Constant ideological agreement may become boring; occasional unexpected and novel content may be welcomed by users. This also helps the user be aware of different views and ideas.

Similarly, better presentation strategies may alert users to the impact of algorithms on the material that is presented to them, and may even allow users to tweak their settings. Additionally, the value of information (e.g. the number of reads by other users) may also be presented more effectively. Better presentation could also nudge people towards their better selves, presenting to them how balanced or biassed their current news reading is and thus encouraging them to check out some alternative views.

Bad personalisation is fraught with potential perils for individuals and society, then. More effective personalisation would address these issues.