Navigating subjectivity in causal context: a non-deterministic approach for information behaviour research

Abstract

Introduction. Information behaviour research often asks causal questions with tools that describe associations. Conventional, regression-based models struggle when constructs are subjective, context-bound, and difficult to estimate. The academic setting is one example where behaviour is often self-directed, cues are fragmentary, social constructs are latent, and perceptions vary across roles and disciplines. Under these conditions, deterministic causes rarely occur and purely probabilistic models can miss how context shapes action.

Method. This paper proposes a hybrid causal procedure—the fuzzy subjective structure model—designed to handle the nuance of human perception. It integrates sufficiency paths (what usually works) from PLS-SEM, bottlenecks (what constrains action) from necessary condition analysis, and the decision-oriented visualization of combined importance–performance map analysis. To address the specific challenges of information behaviour, the model adds two adjustments: it uses entropy weighting to account for how differently people interpret survey questions, and inverse probability weighting to ensure that comparisons between groups (e.g., disciplines) are fair and balanced.

Analysis. The fuzzy subjective structure model formalizes four distinct, non-deterministic causal claims: typical sufficiency (‘if , then typically ’), typical necessity (‘if not , then typically not ’), probabilistic sufficiency (‘if , then probably ’), and probabilistic necessity (‘if not , then probably not ’). This four-claim structure addresses a noted gap in causal analysis for the social sciences by separating ‘works for most cases’ statements from probability-scale, counterfactual-style contrasts.

Results. The fuzzy subjective structure model reorients causal analysis in information research from behaviour-driven explanations to a perception-first assessment with visual aids. It positions latent institutional factors like culture and management as the primary drivers that shape individual behaviour through subjectively perceived affordances and constraints.

Conclusion. The approach supports quantitative inference by identifying conditions that usually enable behaviours and conditions whose absence makes their individual perception unlikely, while keeping latent constructs and small samples in mind.