strategy, as in the worse case, none of these members may be present anymore when the seventh item is shown.
Secondly, overall satisfaction with a sequence may depend more on the order of the items than one would expect. For example, for optimal satisfaction, we may need to ensure that our news program has:
• A good narrative flow. It may be best to show topically related items together. For example, if we have two news items about Michael Jackson (say about his funeral and about a tribute tour) then it seems best if these items are presented together. Similarly, it would make sense to present all sports’ items together.
• Mood consistency. It may be best to show items with similar moods together. For example, viewers may not like seeing a sad item (such as a soldier’s death) in the middle of two happy items (such as a decrease in unemployment and a sporting victory).
• A strong ending. It may be best to end with a well-liked item, as viewers may remember the end of the sequence most.
Similar ordering issues arise in other recommendation domains. For example, a music programme may want to consider rhythm when sequencing items. The recommender may need additional information (such as items’ mood, topics, rhythm) to optimise ordering. It is beyond the topic of this chapter to discuss how this can be done (and is very recommender domain specific). We just want to highlight that the items already shown may well influence what the best next item is. For example, suppose the top four songs in a music recommender were all Blues. It may well be that another Blues song ranked sixth may be a better next selection than a Classical Opera song ranked fifth.
In Experiment 3 (see Figure 21.3), we investigated how a previous item may influence the impact of the next item. Amongst others, we found that mood (resulting from the previous item) and topical relatedness can influence ratings for subsequent items. This means that aggregating individual profiles into a group profile should be done repeatedly, every time a decision needs to be made about the next item to display.