C'est tout complique. Une methode est de simplement remplacer 3 avec un nombre moyen. Puis, pour le reste 1,2,4 et 5, vous savez comment les régler

Thanks for the reply.

I have no idea why you would want to replace the middle value on a Likert-scored item. Also, the currently preferred approach is not to try and guess when someone does or does not have a "response bias." For example, a person who gives all high-scores may indeed feel that way about the overall topic.