Investigating the Impacts of Recommendation Agents on Impulsive Purchase Behaviour

Hui Zhu, Zherui Yang, Carol Xiaojuan Ou, Hongwei Liu, Robert M. Davison

Introduction

In the era of e-commerce, Recommendation Agents (RAs) have become ubiquitous. Whether browsing Amazon, Taobao, or specialized search engines, these algorithms suggest products tailored to individual preferences. While their primary function is to enhance search efficiency, they often lead to a common consumer phenomenon: impulsive purchasing. This study investigates the psychological mechanisms behind how RAs trigger unplanned buying behaviour by reducing cognitive load and social pressure.

Theoretical Framework: The SOR Model

To understand the transition from browsing to impulsive buying, the researchers utilized the Stimulus-Organism-Response (SOR) framework. This model posits that environmental stimuli (S) affect a person's internal state (O), which in turn drives their behavioural response (R).

The conceptual framework based on the Stimulus-Organism-Response (SOR) model.

In this context:

  • Stimulus: RA features including Informativeness, Product Search Effectiveness, and Lack of Sociality Stress.
  • Organism: The consumer's internal psychological state, specifically Perceived Control and Satisfaction.
  • Response: The resulting Impulsive Purchase behaviour.

Methodology

The study analysed data from an online survey of 157 valid respondents in China. The sample was predominantly young (71.4% aged 18-25) and female (58%). Participants reported on their recent experiences with RAs such as Amazon's engine or Taotaosou. The researchers employed Partial Least Squares Structural Equation Modelling (PLS-SEM) to test the hypotheses and ensure the model's statistical validity.

Key Results and Analysis

The structural model successfully explained a significant portion of consumer behaviour. Notably, it accounted for 29.5% of the variance in Impulsive Purchase behaviour and a massive 69.5% of the variance in Satisfaction.

The structural model results showing path coefficients and variance explained (R²).

Key statistical findings include:

  • Informativeness was the strongest predictor of Satisfaction (β = 0.42, p < 0.01).
  • Lack of Sociality Stress significantly enhanced Perceived Control (β = 0.33, p < 0.01).
  • Satisfaction emerged as the primary direct driver of Impulsive Purchase (β = 0.41, p < 0.01).

Interestingly, while Perceived Control was hypothesised to lead directly to impulsive purchases, the data showed its effect was largely mediated or overshadowed by Satisfaction in the full model. Solid lines in the model represent significant paths, while dotted lines indicate paths that did not reach statistical significance.

Discussion: Why RAs Drive Impulses

The "Freedom" of Anonymity

A standout finding is the impact of Lack of Sociality Stress. Traditional retail often involves pushy salespeople who can make shoppers feel obligated or judged. RAs provide a "stress-free" environment where anonymity and the absence of human interaction actually increase a consumer's sense of control and satisfaction, making them more susceptible to impulsive suggestions.

Satisfaction as the Gateway

The research highlights that when an RA provides accurate information and a pleasant experience, the consumer's "guard" drops. High satisfaction levels reduce the perceived risk of an unplanned purchase, effectively acting as a gateway to impulsive spending. This reinforces the "Computers are Social Actors" paradigm, where users respond to algorithms with attitudinal changes similar to human interactions.

Efficiency and Cognitive Load

By improving product search effectiveness, RAs reduce "cognitive overloading." When the machine handles the heavy lifting of filtering and sorting, consumers retain more mental energy to engage emotionally with the shopping experience, which often translates into higher unplanned spending.

Conclusion and Limitations

This study confirms that RAs are more than just functional tools; they are social actors that influence consumer psychology. However, the findings are limited by the cultural context (China) and the reliance on self-reported survey data. Future research could benefit from longitudinal designs or objective behavioural data, such as server logs, to further validate these psychological pathways across different global markets.

Summary updated

BibTeX
@inproceedings{zhu2015investigating,
  author = {Hui Zhu and Zherui Yang and Carol Xiaojuan Ou and Hongwei Liu and Robert M. Davison},
  title = {{Investigating the Impacts of Recommendation Agents on Impulsive Purchase Behaviour}},
  booktitle = {ACIS},
  year = {2015},
}
Figure