A review of an article using argument and questioning methods: “managerial reliance on AI”– for MBA studies students
How to describe the article in Harvard reference format?
The proper Harvard-style reference for the
article based on the provided bibliographic info is:
Cerar, J., 2026. Managerial
reliance on AI in multi-decision contexts: the role of cognitive dissonance and
anticipated workload. Long Range Planning,
59, p.102662.
Highlight the main research issues and research objectives,
2 main academic theories used and the main research methods used in this
article.
Based on the
provided excerpts from the article, here is a summary of the main research
issues, objectives, academic theories, and research methods used:
Main Research
Issues:
·
How managerial reliance on AI
unfolds in complex multi-decision contexts, which involve multiple related
decisions within the same session.
·
The paradoxical tension
managers face between discounting AI advice when they disagree with it, and
relying on AI under conditions of high anticipated workload.
·
The inadequacy of existing
AI-augmentation theories that treat AI reliance in multi-decision contexts as
shaped by individual decisions in isolation, ignoring cognitive dynamics across
decisions.
·
The role of cognitive factors
such as accumulated cognitive dissonance and anticipated workload in shaping
managers' reliance on AI across multiple decisions.
Research Objectives:
·
To develop a path-dependent,
dynamic model explaining how cognitive dissonance and anticipated workload
jointly influence AI reliance over a multi-decision task.
·
To test empirically whether
(1) higher task-level disagreement with AI reduces AI reliance beyond single
decision disagreements, and (2) whether this negative effect is attenuated when
anticipated workload is high.
·
To advance theoretical
understanding of AI-augmented decision-making by linking cognitive dissonance,
workload anticipation, and paradox theory in multi-decision strategic settings.
·
To provide practical guidance
for designing organizational decision processes and AI interfaces that sustain
effective AI integration under multi-decision cognitive demands.
Main Academic
Theories Used:
1. Cognitive Dissonance Theory (Festinger, 1957; Elliot and Devine, 1994):
Extended in this study to multi-decision sessions to explain how repeated
disagreements with AI recommendations generate cumulative cognitive tension
influencing reliance behavior.
2. Workload Anticipation and Paradox Theory: Used to explain how managers navigate the
conflicting demands of cognitive strain (anticipated workload) and distrust
(dissonance) when deciding how much to rely on AI advice in complex tasks.
Main Research
Methods:
·
Field Experiment: Conducted with 88 expert managers evaluating 540
real startup applications in a realistic multi-decision, single-session setting
using an AI-assisted evaluation tool. The experiment manipulated:
·
Task-level disagreement with
AI (high vs. low disagreement by varying AI advice distribution).
·
Anticipated task-level
workload (managers told they would evaluate either 10 or 20 startups, though
all evaluated 10).
·
Managers made initial
evaluations, received AI recommendations, and then had opportunity to revise
their ratings, allowing measurement of AI reliance.
·
Questionnaires to collect demographics, prior AI usage, workload
measures, and psychometric controls.
·
Semi-structured
Interviews (n=11) with a subset of participants
post-experiment, to contextualize quantitative results with qualitative insights
about cognitive experience, perceived AI opacity, and workload effects.
This mixed-methods
approach enabled granular analysis of AI reliance behaviors over multiple
decisions, and links between cognitive dissonance, workload anticipation, and
paradoxical reliance patterns in managerial AI use.
Describe 3 main claims of the article in
terms of Toulmin's model of argument and then provide a critical review of the
article using the Socratic Method.
Below are three
main claims of the article framed using Toulmin’s model of argument, followed
by a critical review employing the Socratic Method.
Three Main Claims Using Toulmin’s Model
Claim 1: Managers’ reliance on AI in multi-decision
contexts is influenced not only by disagreement with AI in single decisions but
also by cumulative disagreement experienced across all decisions in the
session.
·
Grounds (Evidence): Experimental data showing lower AI reliance when
managers experienced higher task-level disagreement across multiple decisions;
interview insights highlighting sustained cognitive tension with repeated
disagreements.
·
Warrant
(Reasoning): Cognitive dissonance theory
predicts cumulative psychological discomfort from repeated exposure to
contradictory information, which influences behavior toward discounting AI
advice.
·
Backing: Established cognitive dissonance literature
(Festinger, 1957; Elliot and Devine, 1994) and advice-taking literature
demonstrating discounting of distant advice (Schultze et al., 2015).
Claim 2: The negative effect of cumulative disagreement
on AI reliance is mitigated when managers anticipate a high cognitive workload
associated with the multi-decision task.
·
Grounds (Evidence): Field experiment manipulation where managers told
to expect 20 decisions (high workload) showed less reduction in AI reliance
despite disagreement, supported by interview statements referencing workload as
a factor affecting reliance.
·
Warrant
(Reasoning): Under high anticipated
workload, managers may rely on AI as a cognitive shortcut or coping mechanism,
balancing the dissonance by accepting AI advice more readily.
·
Backing: Workload anticipation theories (Casper and
Sonnentag, 2020; Casper et al., 2017) and paradox theory explaining navigation
of competing demands without full resolution (Smith and Lewis, 2011).
Claim 3: Effective AI integration in multi-decision
managerial contexts requires considering cumulative cognitive experiences and
task workload, not just technical AI accuracy or interface design.
·
Grounds (Evidence): Practical recommendations arise from findings
showing that repeated disagreement can erode reliance even if AI is accurate;
interviews stressing managerial cognitive strain and opacity concerns .
·
Warrant
(Reasoning): Sustaining trust and
reliance on AI depends on managing cognitive dissonance and workload,
suggesting organizational design implications beyond pure technical focus.
·
Backing: Organizational behavior and human–AI interaction
literature emphasizing the cognitive sustainability of decision-making
processes (Lebovitz et al., 2022; Raisch and Krakowski, 2021).
Critical Review Using the Socratic Method
Question 1: What
is the central problem the article addresses? The article investigates why managers under-utilize AI advice in
multi-decision contexts despite AI’s potential benefits, focusing on cognitive
factors influencing reliance.
Question 2: Does
the article clearly justify the need for studying multi-decision contexts
separately? Yes, it convincingly argues
that single-decision AI reliance theories overlook cumulative cognitive effects
seen in multi-decision sessions, a common real-world scenario .
Question 3: How
strong is the empirical evidence supporting the claims? The evidence is fairly strong given the field
experiment with 88 expert managers evaluating 540 decisions, but with
limitations such as artificial control of task conditions (high vs low
disagreement) which may reduce ecological validity . Small qualitative
sample (11 interviews) offers useful but limited contextualization.
Question 4: Are
alternative explanations for AI reliance behavior adequately considered? The article discusses factors such as AI opacity,
managerial expertise, and risk propensity, but mostly focuses on cognitive
dissonance and anticipated workload. Other psychological or organizational
factors (e.g., emotional responses, team dynamics) could be further explored.
Question 5: Does
the study address variability across managers with different expertise levels? Not fully; the sample involves expert managers,
and the article acknowledges this limitation, suggesting future research on
novice managers to generalize findings.
Question 6: Is the
theoretical integration of cognitive dissonance and paradox theory convincing? Yes, the combination offers a nuanced explanation
of how managers continuously negotiate competing pressures of autonomy and
reliance under cognitive strain, a complex but realistic framing of the
decision process.
Question 7: How
actionable are the practical recommendations? Fairly actionable—suggestions about sequencing tasks, providing
explanations, and managing cognitive load offer concrete steps. However,
implementing these in diverse organizational contexts could be challenging and
warrants tailored approaches.
Question 8: How
could the study be improved? By including
settings with explainable AI capabilities, a broader managerial population, and
more naturalistic multi-decision scenarios with mixed agreement patterns,
future research can increase generalizability and practical relevance.
Summary of Critical Review
The article makes
a valuable theoretical and empirical contribution by revealing cumulative
cognitive dissonance and workload anticipation as key drivers shaping managerial
AI reliance in multi-decision contexts. Its mixed-methods approach strengthens
internal validity and contextual understanding. However, the experimental
design’s constrained conditions and expert-only sample limit ecological
generalizability. The integration of paradox theory enriches the cognitive
dissonance framework, though further exploration of additional psychological
and organizational factors is warranted. Practical recommendations are
insightful but require contextual adaptation. Overall, the study advances
understanding of sustained human-AI interaction amid complex managerial
decision-making demands.
Describe 2 main research limitations of the
article.
·
sting that findings may differ in
settings with more transparent AI.
·
Sample Composed Exclusively of Expert
Managers and Structured Multi-Decision Tasks The study focused on expert managers
performing repeated, structurally similar evaluation tasks within a single
session (e.g., evaluating startups). While this allowed for precise control, it
restricts generalizability. Real-world decision-making often involves more
diverse task types, flexible AI use, and managers with varying expertise
levels. Experts may experience cognitive dissonance differently from
novices—who might trust AI more due to less confidence in their own judgment.
The article acknowledges this limitation and calls for future research on
non-expert managers and more heterogeneous decision-making contexts.
No comments:
Post a Comment