Behavioral Outcomes of Human Cognitive Security within an Integrative Modeling Framework

Abstract

Human decision-making under uncertainty faces growing challenges from information-based threats that pose risks to human cognitive processes and behavior. Although their potential harm is widely acknowledged, there remains no well-defined construct for characterizing the degree to which information-based threats influence changes in human judgments and decision-making, impeding theoretical advancement, measurement, and effective countermeasure development. Here, we introduce a human cognitive security construct focused on linking information-based threats to observable outcomes to bridge field-level definitions with operational measures by drawing from core mechanisms related to information processing and decision-making. To connect the information environment to behavior, we develop an integrative modeling framework that unifies Bayesian inference with affect-modulated decision valuation, capturing how cognitive resource allocation and affective valuation shape three core behavioral outcomes: veracity discernment, task-oriented actions, and information sharing. Through computational simulations, we demonstrate that this framework explains canonical phenomena, including cognitive heuristics, the illusory truth effect (R2=0.86, validated against empirical data), and incongruent veracity discernment and sharing behavior. We propose empirically grounded behavioral outcome measures of cognitive security to guide future empirical examinations. Finally, we outline how environment-specific elements, characterized by data availability and ecological constraints, affect individuals' cognitive security and identify future research directions.

Figures (5)

• Figure 1: From conceptualization to the implementation of a modeling framework, we examine how information-based threats influence human judgment and decision-making under uncertainty toward behavioral outcomes of cognitive security. a. At the core of this study, we examine the influence of information processing on behavior. b. Centrally processed information is mediated by cognitive, affective, and social factors (all regulated by metacognition; hence the dashed line in panel c). Specifically, we model behavioral outputs: queried veracity discernment, directly observable task actions, and directly observable information sharing. c. From left to right, information in the environment must be sensed, perceived, and attended to in order to be processed by the central nervous system towards formulating judgments and decisions. Next, a set of potential hypotheses is considered, and each hypothesis is assigned an evidence-based evaluation, informed by sampled high-level information cues, an allocation of cognitive resources, and the perceived source credibility associated with each piece of information. Following posterior assessment, probabilities of various outcomes associated with a considered action set are computed, often requiring a transition from comprehension to projection. Qualitatively depicted here, judgments and decisions can be further influenced by metacognitive regulation and long-term memory (capturing expertise, training, and recall), which will further dictate the specific descriptive implementation of this modeling framework. Circles notionally represent distinct quantities as they progress through the information processing chain.
• Figure 2: a. A model implementation of a veracity discernment task using the modeling framework presented in Figure \ref{['fig:fig1']}c. b. The normative case where an equal distribution of cognitive resources is allotted, producing a likelihood function that mirrors the underlying uncertainty in the information assessed via an ideal observer (see Methods for detailed implementation). Additionally, this model incorporates unbiased affective valuation of each outcome. c-d. Here, we explore how different cognitive resource mappings produced can describe the emergence of heuristics noted in formulating judgments under uncertainty: availability and anchoring, respectively. Here and throughout Figures \ref{['fig:fig3']} and \ref{['fig:fig4']}, saturated colors indicate where model inputs deviate from the normative case. e. Here, the effect of assigning affective values associated with each action being correct is explored. In this case, selecting 1 being correct is given a disproportional weighting, resulting in a shift in the selection on the veracity discernment task. f. Finally, a lack of perceived source credibility is modeled as a likelihood function that carries no information (as the information is disregarded on the basis of its credibility assessment), resulting in no update of the posterior probability density function from the prior.
• Figure 3: a. A model implementation (using the modeling framework presented in Figure \ref{['fig:fig1']}c) of a veracity discernment task over repeated exposure, producing the illusory truth effect. Computational simulations of the truth effect over information repetitions. b. Truth bias is modeled by biased resource mapping, producing a bias in the likelihood function. c. The emergence of illusory truth from truth bias in a likelihood function is demonstrated by plotting the mean of the prospective value (action selection) overlaid with empirical data hassan_effects_2021.
• Figure 4: a A model implementation (using the modeling framework presented in Figure \ref{['fig:fig1']}c) of a share / no share paradigm (see Methods for the specific implementation). b. Normative sharing behavior, where erroneous information is correctly identified and not shared due to proportional benefits and losses associated with sharing veridical and erroneous information, respectively. c. Here, misaligned sharing behavior is described when the value of social engagement associated with sharing potentially false information outweighs the value of not sharing information. d. Full vulnerability to the information-based threat is realized as erroneous information is not correctly identified due to the allocation of cognitive resources along the hypothesis space, despite proportional benefits and losses associated with sharing veridical and erroneous information.
• Figure 5: a A depiction of how ecological constraints, such as time pressure, can result in only a subset of information that can be processed by the human, denoted as utilizable information. b. The information processing spectrum of decision-making conceptually presented and categorized into low information processing (black outline) and high information processing environments (white outline), given shared elements expected to modulate judgments and decision-making, thus vulnerability to information-based threats. The information processing spectrum incorporates the amount of data available in the environment as well as ecological constraints on operators. Notional contour lines illustrate constant information across varying ecological constraints and data availability. On the most extreme end of low information processing environments (upper left corner), decreased information access and high ecological constraints largely coincide with operations in ICE environments. On the extreme end of high information processing environments (lower right corner), increased information access and low ecological constraints coincide with interactions in social media environments. Of importance for cognitive security are transitional groups (center; dashed outline), where information-based threats in one domain could spill over into impaired behavioral aspects in another domain. Notional naturalistic operational roles are assigned across the information processing spectrum.