Embodied Pedagogical Agents From Visual Impact to Pedagogical Implications

Embodied pedagogical agents – v sually represented, computer generated characters n pedagog cal roles, such as v rtual nstructors, mentors and learn ng compan ons – populate the d g tal soc ety n ncreas ng numbers. They are found n educat onal programmes from preschool to un vers ty, as well as n broader educat onal contexts n the form of med cal counsellors, v rtual health coaches, nformat on gu des, etc. S nce long, they also appear n eduta nment and nfota nment sett ngs. The thes s addresses psycholog cal, cogn t ve and soc al aspects of embod ed v rtual characters n pedagog cal contexts w th respect to ‘stat c v sual character st cs’, .e. underly ng v sual character st cs. On the bas s of theoret cal cons derat ons as well as several emp r cal stud es, t s argued that users’ v sual and aesthet c exper ence of embod ed pedagog cal characters (EPCs) s too mportant to be treated as a secondary ssue – not the least w th respect to central pedagog cal goals to mot vate and engage. The ma n conclus ons of the thes s are: ( ) that v sual aspects n EPCs can be related to pedagog cal outcomes, ( ) that there may be s gn ficant relat ons between v sual and soc al character st cs of EPCs and learner character st cs, and ( ) that EPCs may reproduce stereotypes from everyday real l fe human-human nteract on, as well as from trad t onal v sual med a – but that they s multaneously harbour a cons derable potent al to challenge stereotypes. As a tool for the research commun ty, a framework of a v sual graph cal des gn space s proposed. The framework may scaffold the des gn of user evaluat ons, wh ch often are blurred by uncontrolled v sually related var ables. W th regard to weak user evaluat ons, the thes s furthermore argues that there s a problem w th user evaluat ons based upon the concept of ‘the User’, .e. an averaged standardzed user. In order to reveal mportant correlat ons between agent character st cs and user character st cs, t s often necessary to dent fy relevant subgroups of users rather than to treat the whole user group as an ‘averaged user’. And finally – v rtual agents may have a not yet thought of potent al as research