ABSTRACT 

The rapid proliferation of conversational agents powered by large  language  models  presents  a  fundamental  challenge  to  human-AI  interaction:  users  systematically  vary  in  how  they  perceive  and  interact with these systems, yet no coherent theoretical framework  explains  these  patterns.  This  paper  proposes  a  Neo-Piagetian  developmental  framework  to  understand  anthropomorphic  response patterns in conversational agent interactions. Drawing on  domain-specific  cognitive  development  theory,  we  demonstrate  that  users  progress  through  distinct  developmental  stages— preoperational,  concrete  operational,  and  formal  operational— characterized by qualitatively different ways of understanding AI  systems.  Through  conceptual  framework  methodology,  we  synthesize  empirical  evidence  showing  how  users  with  low  AI  literacy  exhibit  animistic  thinking  and  egocentrism  (preoperational),  while  those  with  moderate  literacy  develop  functional classifications without grasping underlying mechanisms  (concrete  operational),  and  high-literacy  users  achieve  abstract  reasoning about computational architectures (formal operational).  We introduce a cyclical model of AI literacy development driven  by  Piagetian  mechanisms  of  assimilation,  disequilibrium,  and  equilibration,  where  unexpected  AI  behaviors  trigger  cognitive  conflict  that  propels  developmental  progression.  This  framework  resolves  contradictions  in  existing  literature  by  showing  that  anthropomorphic  responses  reflect  users'  developmental  stage  rather  than  random  variation,  and  that  the  same  individual  can  exhibit different cognitive levels across domains. The framework  has  significant  implications  for  AI  design,  deployment,  and  education,  suggesting  that  systems  should  adapt  to  users'  developmental stages and that literacy interventions can accelerate  progression  beyond  primitive  anthropomorphic  thinking.  As  conversational  agents  become  increasingly  sophisticated  in  mimicking  human  communication,  understanding  these  developmental  dynamics  becomes  critical  for  designing  systems  that balance accessibility with appropriate user understanding.