Text-Based Simulation of Human-Computer Interaction Using Cognitive-Affective Architectures and Natural Language Interfaces

Abstract

This study presents a text-based simulation framework for modeling Human-Computer Interaction (HCI) using a hybrid cognitive-affective architecture integrated with natural language processing tools. Rather than relying on external corpora, we generate a synthetic dataset of 27 text samples that reflect diverse emotional and linguistic patterns commonly observed in intelligent systems' interactions. Our methodology employs symbolic and statistical analysis, combining sentiment metrics (polarity, subjectivity) with structural linguistic features (word count, part-of-speech distribution, and average word length). Each sentence acts as a micro-interaction unit, evaluated through a custom Python pipeline using TextBlob, TF-IDF, cosine similarity, and visualization libraries. The simulation reveals that textual interactions, when modeled with both emotional and cognitive dimensions, can mimic realistic communication patterns, adapt to affective cues, and provide a platform for dialogue system benchmarking. A similarity matrix derived from cosine distances further supports the clustering of thematically aligned interactions, demonstrating the system’s interpretive fidelity. Results indicate that the proposed hybrid model captures subtle emotional nuances and structural variations across user expressions. This dual-layered modeling opens avenues for more emotionally intelligent dialogue agents and enables refined evaluation in early-stage HCI development environments.