At the expense of statistical power and generalisability, you could oversample tiramisu (e.g., 80% of all assessments vs. 20% randomly allocated to a selection of other foods).

Is this based on correlational within-person data or are you using state-of-the-art N = 1 experimental designs?

We end with a call to action for the collection of more multi-methodological and behavioral data on personality. In doing so, we outline key future directions to go beyond the sole use of self-report in a truly behavioral personality science.

Interestingly, self-reported states (e.g., from experience sampling) were rarely incrementally predictive above classical global self-report and thus not as close to behavior as commonly believed. In turn, other-reported states were predictive beyond global informant-report and self-reported states.

We provide an initial examination in a uniquely multi-methodological study (N = 88) and examine how six trait measures (global self- and informant-report, event-based and time-based self-reported as well as peer-reported states) predict actual behavior observed in the laboratory one year later.

Personality research on behavioral traits typically focuses on global self-report. However, alternatives exist, concerning (1) type of aggregation (global reports vs. average state reports) and (2) perspective (self-report vs. other-report), and have been suggested to be closer to behavior.

I'd love to be added as well! :)