Recommendation engines:

Transparency and transferability

The value chain of recommendation (pictured below) is somewhat distorted.

The recommandation happens solely on the terms of the channel owner while producers as well as users are left with none or very limited possibilities to affect the outcome, i.e. there is a clear lack of transparency  and the dataflow is strictly one-directional.



By editing their own preference information the users can do some changes, but this does not get one far as the recommendation is usually built on historical data and collaborative filtering-methods.

The user preference profile is also channel-specific and untransferrable, which means that the users need to start building a profile from scratch when switching channels.

Producers can affect the recommendation solely by delivering product and service related metadata to the channel, although the incentives to do this are low due to the one-directional dataflow.

Great value and insights are missed  as producers lack the chance to understand, learn and act on users preferences.


the GOAL

To conceptualize a new transparent recommendation engine enabling transferability of user preference profiles (pictured below).

User: The ability to edit preference information and transfer preference profiles between channels.

Producers:  Can easily produce product-specific metadata enabling better product recommendation. The producers also get information about user preferences.

Channel: Transferable profiles from other channels make the recommendation easier and faster.


Two stages:

1. The transferability is built within an organisations having various own channels.

2. The transferability is built between channels from different organisations.


Producers, channel owners and service providers.

Finnish AI Accelerator

contact information

00130 Helsinki

Finnish AI Accelerator Tekoälyaika
Finnish AI Accelerator Teknologiateollisuus ry