If we move to systems that provide recommendations based on previous preferences and behaviour, how can we also achieve the chance of serendipity?
Recommendation engines and AI can often help with serendipitous discovery by revealing connections a user wouldn’t otherwise know exists. It depends on how they are implemented, but they can break the reader out of their comfort zone by recommending pieces of content that they were not expecting. This may then encourage them to explore further.
Humans typically search based on what they know or what they expect e.g. within particular journals, in specific fields or using industry keywords. They wouldn’t typically think to search literature from a different industry, domain etc. An AI driven recommendation engine can do this for the user. It is interesting to note that cross functional and cross industry research is what is needed to solve some of the biggest problems we face (eg. climate change) and an AI approach can help with this.
Imagine if someone could share their interests with platforms automatically when they sign in.
It is more appropriate to infer users’ interests from their behaviour. For example, looking at what they have published previously, the search terms they use, or the research papers they look at. This gives a much richer and more detailed view of what a user is really interested in and allows us to provide a much better personalised service to them, that evolves over time, alongside their needs. Asking them to state their interest in very broad areas upfront only takes a snapshot – and doesn’t generate a true picture of the multiple different concepts they are interested in over time.