Institut für Business Analytics

Design and Evaluation of Explanations for
User-Centric Human–AI Interactions

Philipp Schröppel, 14 September 2026
Universität Ulm

Agenda

1
Overview
Motivation, Publications, Methodology
2
Deep Dive
Choose Wisely: Leveraging Explainable AI to Support Reflective Decision-Making, ICIS 2024
3
?
Discussion
Q&A
2

Modern AI Has Rapidly Grown in Scale and Capability

2012
AlexNet
ImageNet winner
2022
ChatGPT (GPT-3.5)
Fastest-growing consumer app
2026
GPT-6 Astra
“Anything you can do on a computer, Astra can do for you.”
60M parameters
1.9 years
175B parameters
5,500 years
10T parameters
317,100 years
Model size — time to count the parameters, one per second
3

The Same Conclusion Can Result From Transparent Rules or Opaque Learned Representations

If it rained, the street is wet.  The street is not wet.
p → q ,  ¬q  ⊢  ¬p
Rule-based
Six rules, written by hand
ONE STEP = ONE APPLIED RULE
if rained → wet← premise
not wet← premise
if not wet → not rained← contraposition
∴ not rained← modus ponens
It did not rain.
Learning-based
Billions of parameters, learned from data
ONE STEP = ONE PREDICTED WORD
If it rained , the street is wet . The street is not wet . It did ?
not0.61
probably0.22
certainly0.07
It did not rain.
4

Opacity Blocks Responsible Decision-Making

EXAMPLE
A recruiter decides on candidates for a position and an opaque AI system recommends rejecting candidate Alice.
WITHOUT AI SUPPORT
WITH AI SUPPORT
ATTRIBUTABILITY 
Does the decision reflect the decision-maker’s judgment?
Can the grounds for the decision be critically evaluated?
Why do I consider candidate Alice unsuitable?
Can the grounds underlying the AI recommendation be critically evaluated?
Why does the AI consider candidate Alice unsuitable?
ACCOUNTABILITY 
Is the decision-maker answerable to others?
Can the decision be justified to affected stakeholders?
Can I justify the rejection to candidate Alice?
Can the decision be justified when part of its grounds is inaccessible?
Can I justify the rejection to candidate Alice?
Opaque AI weakens the basis for responsible judgment without reducing responsibility for the decision.
5

Improving Human–AI Interaction Requires Designing and Empirically Evaluating Explanations

Opacity Modern AI gives an output but not its grounds, so the decision-maker can neither evaluate nor justify it.
Explanation An explanation answers a why-question by communicating the causes of an outcome.
Design Which causes count, and how they must be framed, depends on the recipient and the situation. Explanations have to be designed for their intended use.
Evaluation Whether an explanation is effective only shows in how people interact with it. Explanations have to be evaluated empirically with users.
Research Objective (RO)
To improve human–AI interaction through the design and evaluation of explanations.
6

The Dissertation Comprises Six Papers Across Leading IS Outlets

#
SHORT TITLE
AUTHORS
OUTLET
RANK*
STATUS
1
Exploring XAI Users’ Needs
P. Schröppel, M. Förster
European Conference on Information Systems
A
Published 2024
2
Navigating the Rashomon Effect
J. Rosenberger, P. Schröppel, S. Kruschel, M. Kraus, P. Zschech, M. Förster
European Conference on Information Systems
A
Published 2025
3
User-Centric Chain-of-Thought Reasoning
P. Schröppel
Hawaii International Conference on System Sciences
B
Published 2026
4
Let’s Chat About Overreliance
K. Pitz, P. Schröppel, C. Tille, K. Züllig, S. Zimmermann
Internationale Tagung Wirtschaftsinformatik
B
Published 2026
5
Choose Wisely
M. Förster, P. Schröppel, C. Schwenke, L. Fink, M. Klier
International Conference on Information Systems
A
Published 2024
6
Co-Creation to Enhance Reflective Decision-Making
M. Förster, P. Schröppel, C. Schwenke, M. Klier, L. Fink
European Journal of Information Systems
A
Submitted 2026
*VHB-Rating 2024 Wirtschaftsinformatik — vhbonline.org
7

Human–AI Interactions Range From Receiving Outputs to Contesting and Shaping Them

LESS
USER AGENCY
MORE
AI
OUTPUT
USER
TOPIC & RESEARCH OBJECTIVES
Receiving
Accept or reject a finished output
TOPIC I
Personalization of AI explanations
RO1
To design and evaluate a personalization approach for post-hoc AI explanations.
RO2
To design and evaluate a personalization approach for intrinsically interpretable models.
Contesting
Probe and push back on its reasoning
TOPIC II
Scrutinization of LLM outputs
RO3
To design and evaluate a user-centric approach to chain-of-thought reasoning that makes LLM reasoning traces structured, inspectable, and correctable by users.
RO4
To understand how AI literacy training and AI explanations jointly shape appropriate reliance on LLM-based AI systems.
Shaping
Feed inputs into how it is generated
TOPIC III
Reflective decision-making with AI
RO5
To design and evaluate an interactive AI explanation approach to foster reflective decision-making.
RO6
To empirically investigate how co-creation of AI outputs shapes reflective decision-making with AI.
8

Human–AI Interactions Range From Receiving Outputs to Contesting and Shaping Them

LESS
USER AGENCY
MORE
AI
OUTPUT
USER
TOPIC & RESEARCH OBJECTIVES
Receiving
Accept or reject a finished output
TOPIC I
Personalization of AI explanations
RO1
To design and evaluate a personalization approach for post-hoc AI explanations.
RO2
To design and evaluate a personalization approach for intrinsically interpretable models.
Contesting
Probe and push back on its reasoning
TOPIC II
Scrutinization of LLM outputs
RO3
To design and evaluate a user-centric approach to chain-of-thought reasoning that makes LLM reasoning traces structured, inspectable, and correctable by users.
RO4
To understand how AI literacy training and AI explanations jointly shape appropriate reliance on LLM-based AI systems.
Shaping
Feed inputs into how it is generated
TOPIC III
Reflective decision-making with AI
RO5
To design and evaluate an interactive AI explanation approach to foster reflective decision-making.
RO6
To empirically investigate how co-creation of AI outputs shapes reflective decision-making with AI.
8

Human–AI Interactions Range From Receiving Outputs to Contesting and Shaping Them

LESS
USER AGENCY
MORE
AI
OUTPUT
USER
TOPIC & RESEARCH OBJECTIVES
Receiving
Accept or reject a finished output
TOPIC I
Personalization of AI explanations
RO1
To design and evaluate a personalization approach for post-hoc AI explanations.
RO2
To design and evaluate a personalization approach for intrinsically interpretable models.
Contesting
Probe and push back on its reasoning
TOPIC II
Scrutinization of LLM outputs
RO3
To design and evaluate a user-centric approach to chain-of-thought reasoning that makes LLM reasoning traces structured, inspectable, and correctable by users.
RO4
To understand how AI literacy training and AI explanations jointly shape appropriate reliance on LLM-based AI systems.
Shaping
Feed inputs into how it is generated
TOPIC III
Reflective decision-making with AI
RO5
To design and evaluate an interactive AI explanation approach to foster reflective decision-making.
RO6
To empirically investigate how co-creation of AI outputs shapes reflective decision-making with AI.
8

Six Research Objectives Address Three Topics in Human–AI Interaction

#
SHORT TITLE
TOPIC
RESEARCH OBJECTIVE
1
Exploring XAI Users’ Needs
Topic I — Personalization of AI explanations
RO1 — To design and evaluate a personalization approach for post-hoc AI explanations.
2
Navigating the Rashomon Effect
RO2 — To design and evaluate a personalization approach for intrinsically interpretable models.
3
User-Centric Chain-of-Thought Reasoning
Topic II — Scrutinization of LLM outputs
RO3 — To design and evaluate a user-centric approach to chain-of-thought reasoning that makes LLM reasoning traces structured, inspectable, and correctable by users.
4
Let’s Chat About Overreliance
RO4 — To understand how AI literacy training and AI explanations jointly shape appropriate reliance on LLM-based AI systems.
5
Choose Wisely
Topic III — Reflective decision-making with AI
RO5 — To design and evaluate an interactive AI explanation approach to foster reflective decision-making.
6
Co-Creation to Enhance Reflective Decision-Making
RO6 — To empirically investigate how co-creation of AI outputs shapes reflective decision-making with AI.
9

IS Research Connects Real-World Socio-Technical Problems With Scientific Knowledge Through Design Science Research (DSR) and Behavioral Science Research (BSR)

ISR FRAMEWORK — ADAPTED FROM HEVNER ET AL. 2004
Environment
Relevance
IS research
Rigor
Knowledge Base
Business Needs
Application to the Environment
Applicable Knowledge
Additions to the Knowledge Base
People
Roles
Capabilities
Characteristics
Organizations
Strategies
Structure and culture
Processes
Technology
Infrastructure
Applications
Architecture
Build / Develop
DSR
Build artifacts
BSR
Develop Theories
Assess / Refine
Research Outcomes
Evaluate / Justify
DSR
Evaluate Artifacts
BSR
Justify Theories
Simulation,  Experiment, 
Field Study
Foundations
Theories
Frameworks
Constructs
Models
Methods
Instantiations
Methodologies
Data Analysis Techniques
Formalisms
Measures
Validation Criteria
10

Explanation Evaluation Ranges From Formal Proxies to Real Users Performing Real Tasks

INTERPRETABLE MACHINE LEARNING EVALUATION TAXONOMY, ADAPTED FROM  DOSHI-VELEZ AND KIM 2017
MORE SPECIFIC AND COSTLY
EVALUATION TIER
HUMANS
TASKS
Application-Grounded Evaluation
Domain experts on the deployed task, in context
Real
Humans
Real
Tasks
Human-Grounded Evaluation
Lay participants on a simplified proxy of the task
Real
Humans
Simple
Tasks
Functionally-Grounded Evaluation
A formal proxy for interpretability, no human subjects
No Real
Humans
Proxy
Tasks
11

The Studies Combine DSR and BSR With Increasingly Realistic Evaluation

#
SHORT TITLE
PARADIGM / TIER
TASK
MODEL / EXPLANATION
METHOD / DATA
1
Exploring XAI Users’ Needs
DSR / Human-grounded
Image classification
CNN / Feature importance
Online experiment / interaction data
2
Navigating the Rashomon Effect
DSR / Functionally- & human-grounded
Demand forecasting
GAM / Feature shape plots
Online experiment / system + interaction data
3
User-Centric Chain-of-Thought Reasoning
DSR / Functionally- & human-grounded
Math word problem solving
LLM / Chain-of-thought traces
Offline benchmarking + online experiment
4
Let’s Chat About Overreliance
BSR / Human-grounded
Hate speech detection
LLM / Citations, references
Online experiment / interaction data
5
Choose Wisely
DSR / Human-grounded
Educational decision-making
Text embedding / Concept explanations
Online experiment / survey
6
Co-Creation to Enhance Reflective Decision-Making
BSR / Application-grounded
Educational decision-making
Text embedding / Concept explanations
Online field experiment / interaction + survey
12

Recap: Six Papers, Three Topics, Six Research Objectives, Two Paradigms

#
SHORT TITLE
OUTLET
RANK*
STATUS
TOPIC
RO
PARADIGM
1
Exploring XAI Users’ Needs
European Conference on Information Systems
A
Published 2024
I
RO1
DSR
2
Navigating the Rashomon Effect
European Conference on Information Systems
A
Published 2025
I
RO2
DSR
3
User-Centric Chain-of-Thought
Reasoning
Hawaii International Conference on System Sciences
B
Published 2026
II
RO3
DSR
4
Let’s Chat About Overreliance
Internationale Tagung Wirtschaftsinformatik
B
Published 2026
II
RO4
BSR
5
Choose Wisely
International Conference on Information Systems
A
Published 2024
III
RO5
DSR
6
Co-Creation to Enhance
Reflective Decision-Making
European Journal of Information Systems
A
Submitted 2026
III
RO6
BSR
*VHB-Rating 2024 Wirtschaftsinformatik — vhbonline.org
13

Agenda

1
Overview
Motivation, Publications, Methodology
2
Deep Dive
Choose Wisely: Leveraging Explainable AI to Support Reflective Decision-Making, ICIS 2024
3
?
Discussion
Q&A
14

Critical Decisions Require AI That Fosters Rather Than Replaces Reflection

Reflection: critical analysis of information and experience to inform decisions.
Reflection-in-action: real-time reflection during the activity — a “conversation with the situation”.
AI can induce reflection when it augments rather than substitutes human decision-making.
Efficiency-focused AI is harmful in critical contexts: hasty acceptance yields sub-optimal, irreversible decisions. Facing a black box, users can only follow blindly or disregard.
Educational choices are non-routine with long-term impact.
RESEARCH OBJECTIVE 5
To design and evaluate an interactive AI explanation approach to foster reflective decision-making.
DSR PROCESS
Peffers et al. 2007
1
Problem identification
2
Objectives of a solution
3
Design and development
4
Demonstration
5
Evaluation
6
Communication
15

AI for Reflective Decision-Making Must Support Five Capabilities

Exploration
Investigate a wider set of educational alternatives
Self-reflection
Examine one’s own thoughts, priorities and behavior
Autonomy
Decide on one’s own judgment instead of accepting advice
Confidence
Belief in one’s ability to decide well-informedly
Trust
Calibrated reliance on the recommender
DSR PROCESS
Peffers et al. 2007
1
Problem identification
2
Objectives of a solution
3
Design and development
4
Demonstration
5
Evaluation
6
Communication
16

A Concept-Bottleneck Model Based on Text Embeddings

USER PROFILE
01Concept relevance
02Course ranking
{{ edges }}
User Profile Business Administration Statistics
Digital Business & Analytics {{ sa1 }}
Market Analysis & Machine Learning {{ sa2 }}
Strategy & Corporate Finance {{ sa3 }}
External Accounting {{ sa4 }}
Sustainable Corporate Governance {{ sa5 }}
1
2
3
4
5
6
Design and Development
17

Design Step 1: Concepts as a Shared Language

USER PROFILE
01Concept relevance
02Course ranking
{{ edges }}
Concepts as a Shared Language
User Profile Business Administration Statistics
Digital Business & Analytics {{ sa1 }}
Market Analysis & Machine Learning {{ sa2 }}
Strategy & Corporate Finance {{ sa3 }}
External Accounting {{ sa4 }}
Sustainable Corporate Governance {{ sa5 }}
1
2
3
4
5
6
Design and Development
18

Design Step 2: Concept-Based Explanations

USER PROFILE
01Concept relevance
02Course ranking
{{ edges }}
Concept-Based Explanations
User Profile Business Administration Statistics
Digital Business & Analytics {{ sa1 }}
Market Analysis & Machine Learning {{ sa2 }}
Strategy & Corporate Finance {{ sa3 }}
External Accounting {{ sa4 }}
Sustainable Corporate Governance {{ sa5 }}
1
2
3
4
5
6
Design and Development
19

Design Step 3: Concept-Based Interventions

USER PROFILE
01Concept relevance
02Course ranking
{{ edges }}
Concept-Based Interventions
User Profile Business Administration Statistics
Digital Business & Analytics {{ sa1 }}
Market Analysis & Machine Learning {{ sa2 }}
Strategy & Corporate Finance {{ sa3 }}
External Accounting {{ sa4 }}
Sustainable Corporate Governance {{ sa5 }}
1
2
3
4
5
6
Design and Development
20

Interventions alter recommendations by acting on concepts

USER PROFILE
01Concept relevance
02Course ranking
{{ edges }}
User Profile Business Administration Statistics
Digital Business & Analytics {{ sa1 }}
Market Analysis & Machine Learning {{ sa2 }}
Strategy & Corporate Finance {{ sa3 }}
External Accounting {{ sa4 }}
Sustainable Corporate Governance {{ sa5 }}
1
2
3
4
5
6
Design and Development
21
ai-scouty.skill-kompass.de

The Experiment Isolates the Effects of Explanation and Intervention

PROCEDURE
Data input
Round 1
Continue
Leave
Round 2
Continue
Leave
Round 3
Leave
Overview
GROUP 1 — CONTROL
n = 150
Recommendation
Concept Explanation
Concept Intervention
GROUP 2 — EXPLANATION
n = 184
Recommendation
Concept Explanation
Concept Intervention
GROUP 3 — FULL ARTIFACT
n = 161
Recommendation
Concept Explanation
Concept Intervention
SAMPLE
N = 495
of 591 recruited on Prolific (March 2024), 96 removed for failed attention checks, students of business-related majors, fluent in English
279 female / 214 male / 2 not stated, 56 countries of origin, most frequently South Africa (182), UK (53), Poland (38), ~10 min per participant
1
2
3
4
5
6
Evaluation
23

Explanations Increase Reflective Decision-Making When Combined With Interventions

Exploration
ANOVA: F = 6.461   p = 0.002
1.0
2.0
3.0
4.0
5.0
6.0
7.0
**
***
5.60
5.16
5.22
Explanation &
Intervention
Explanation
Control
Self-reflection
ANOVA: F = 3.930   p = 0.020
1.0
2.0
3.0
4.0
5.0
6.0
7.0
*
**
5.98
5.65
5.72
Explanation &
Intervention
Explanation
Control
Autonomy
ANOVA: F = 1.477   p = 0.229
1.0
2.0
3.0
4.0
5.0
6.0
7.0
5.55
5.32
5.39
Explanation &
Intervention
Explanation
Control
Confidence
ANOVA: F = 5.516   p = 0.004
1.0
2.0
3.0
4.0
5.0
6.0
7.0
*
**
5.78
5.41
5.49
Explanation &
Intervention
Explanation
Control
Trust
ANOVA: F = 4.215   p = 0.015
1.0
2.0
3.0
4.0
5.0
6.0
7.0
*
**
5.68
5.36
5.39
Explanation &
Intervention
Explanation
Control
Usefulness
ANOVA: F = 5.573   p = 0.004
1.0
2.0
3.0
4.0
5.0
6.0
7.0
**
**
6.06
5.73
5.70
Explanation &
Intervention
Explanation
Control
* p<0.05, ** p<0.01, *** p<0.001 (LSD test). Bars show means, whiskers ± 1 standard error.
1
2
3
4
5
6
Evaluation
24

Current Limitations Point to Broader and More Realistic Evaluations

Limitation
Future research
1
Single use case — generalization to other domains untested
Apply and evaluate the approach in other domains and with other target groups
2
Human-grounded evaluation only; participants were not personally affected by their decisions
Move to application-grounded evaluation, e.g. field experiments
3
Subjective measures dominate — besides exploration, no objective data
Add objective measures such as completion rates, plus qualitative methods
4
Concept extraction was a first pass only
Let users contribute their own personally relevant concepts
1
2
3
4
5
6
Communication
25

Agenda

1
Overview
Motivation, Publications, Methodology
2
Deep Dive
Choose Wisely: Leveraging Explainable AI to Support Reflective Decision-Making, ICIS 2024
3
?
Discussion
Q&A
26

The Capital Committed to AI

ANNUAL AI INFRASTRUCTURE CAPEX. USD BN
0 400 800 1200 1600 $765 $1,011 $1,220 $1,392 $1,579 $1,636 German federal budget 2026 — $635 bn 2026 2027 2028 2029 2030 2031
Compute
Data centers
Power
Source: G. Lee & L. Greenbaum (2026), Tracking Trillions: The Assumptions Shaping the Scale of the AI Build-Out, Goldman Sachs Global Institute. Baseline anchored to NVIDIA data-center revenue estimates. German federal budget: Bundesministerium der Finanzen, Bundeshaushalt 2026, €545 bn, converted at €1 = $1.165 (10 Sep 2026).
27

Where That Electricity Is Consumed

DATA-CENTER ELECTRICITY CONSUMPTION, TWh PER YEAR
0 200 400 600 800 1000 415 TWh ≈ 945 TWh Germany — total electricity generation 2025 — 509 TWh 2024 2030 base case
United States
China
Europe
Rest of world
Source: IEA (2025), Data centre electricity consumption by region, Base Case, 2020-2030, IEA, Paris. https://www.iea.org/data-and-statistics/charts/data-centre-electricity-consumption-by-region-base-case-2020-2030, Licence: CC BY 4.0. Regional values derived from the published 2024 shares and 2024–2030 growth figures in Energy and AI. German reference: Destatis, gross electricity production, 509.2 TWh.
28

Paper 5 - Construct Items

Choose Wisely — ICIS 2024
Exploration
Sun et al. (2023)
–AI Scouty found me a broad set of courses.
–AI Scouty found courses for me that offered a lot of variety.
–AI Scouty found interesting courses for me that I had not considered before.
–AI Scouty found courses for me that covered areas where I want to make progress.
Self-reflection
Grant et al. (2002)
–AI Scouty has helped me to think about which course I would like to take.
–AI Scouty has helped me to think about what areas of interest I want to progress in.
Confidence
Taylor and Betz (1983)
–I am confident that with the help of AI Scouty, I have at least found one course that suits my interests.
–I am confident that with the help of AI Scouty, I would not worry about whether the course fits my interests or not.
–When interacting with AI Scouty, I get information about courses I am interested in.
–I am confident that with the help of AI Scouty, I can figure out what an ideal course would be.
Autonomy
Sankaran et al. (2021)
–I really enjoyed this way of finding courses.
–I think I would actively use AI Scouty during my studies to find courses.
–AI Scouty allowed me to act when the course recommendation was not a good fit for me.
–AI Scouty let me keep control of the course recommendations.
Trust
Benbasat and Wang (2005)
–AI Scouty has the expertise to understand my needs and preferences about courses.
–AI Scouty has good knowledge about courses.
–AI Scouty puts my interests first.
–AI Scouty wants to understand my needs and preferences.
–AI Scouty provides unbiased course recommendations.
Usefulness
Benbasat and Wang (2005)
–Using AI Scouty would improve reflective decision-making on which course to take.
–Using AI Scouty would make it easier to make reflective decisions about which course to select.
–Using AI Scouty would enhance my effectiveness in assessing a subset of individually relevant courses based on areas such as interests.
–Using Scouty would make it easier to find suitable courses.
29

Paper 5 - Distribution of Drop-Out Rates

Choose Wisely — ICIS 2024
PARTICIPANTS LEAVING (% WITHIN GROUP) EXPLANATION & INTERVENTION EXPLANATION CONTROL
After first round 4.9% 13.2% 11.9%
After second round 21.5% 17.4% 12.6%
After third round 73.6% 69.5% 75.5%
30

Paper 1 — XAI Personalization Approach

Exploring XAI Users’ Needs — ECIS 2024
Five-step XAI personalization loop
31

Paper 1 — Algorithm

Exploring XAI Users’ Needs — ECIS 2024
Algorithm: XAI personalization approach
32

Paper 1 — Study Task

Exploring XAI Users’ Needs — ECIS 2024

Screenshot of the study interface
33

Paper 1 — Results

Exploring XAI Users’ Needs — ECIS 2024

Hyperparameter distributions, reward-model variance, and mean effectiveness ratings
34

Paper 2 — Approach

Navigating the Rashomon Effect — ECIS 2025
Six-step interpretability personalization loop
35

Paper 2 — Model Variants

Navigating the Rashomon Effect — ECIS 2025
Progression of shape-function sets across configurations
36

Paper 2 — Reward Distributions

Navigating the Rashomon Effect — ECIS 2025
Histograms of mean reward by hyperparameter level
37

Paper 2 — Learning Behavior

Navigating the Rashomon Effect — ECIS 2025
Reward-model variance per round and weights against cumulative reward
38

Paper 3 — Artifact

User-Centric Chain-of-Thought Reasoning — HICSS 2026
Tagged reasoning output and its rendered user-facing trace
39

Paper 3 — Results

User-Centric Chain-of-Thought Reasoning — HICSS 2026

Construct means with standard errors and box plots by condition
Ease of use and usefulness differ significantly (*); trust is comparable across conditions.
40

Paper 4 — Research Model

Let’s Chat About Overreliance — WI 2026

Research model: AI literacy training and explanation act through engagement with the chatbot and confidence in the AI prediction on overreliance
Two manipulated factors (H1–H4), two mediators, one outcome — overreliance (H5, H6).
41

Paper 4 — Experimental Procedure

Let’s Chat About Overreliance — WI 2026

Procedure: introduction, random assignment to AI literacy training, task description, random assignment to task with or without explanation, survey
Two independent randomisations — training (2a / 2b) and explanation (4a / 4b).
42

Paper 4 — Engagement with the Chatbot

Let’s Chat About Overreliance — WI 2026

Number of User Queries Mean Message Length
(1) (2) (3) (4)
Training 0.360**
(0.182)		 0.473*
(0.256)		 60.171***
(20.320)		 46.723*
(27.960)
Explanation -0.296
(0.181)		 -0.175
(0.262)		 -110.371***
(20.252)		 -126.073***
(30.184)
Training × Explanation -0.231
(0.364)		 28.555
(40.714)
Constant -1.122***
(0.164)		 -1.179***
(0.189)		 410.344***
(17.851)		 417.751***
(20.754)
Observations 1,068 1,068 411 411
Cond./Marg. R2 0.492/0.025 0.493/0.026 0.331/0.133 0.331/0.134
Notes: ***p < 0.01, **p < 0.05, *p < 0.1. Standard errors in parentheses.
Paper 4 research model
43

Paper 4 — Confidence in the AI

Let’s Chat About Overreliance — WI 2026

Confidence in AI
(5) (6)
Number of User Queries -0.463***
(0.059)		 -0.496***
(0.099)
Mean Message Length -0.002***
(0.0004)
Constant 6.079***
(0.066)		 7.057***
(0.221)
Observations 1,068 411
Cond./Marg. R2 0.209/0.064 0.382/0.129
Notes: ***p < 0.01, **p < 0.05, *p < 0.1. Standard errors in parentheses.
Paper 4 research model
44

Paper 4 — Overreliance

Let’s Chat About Overreliance — WI 2026

Overreliance
(7) (8)
Confidence in AI 0.202**
(0.046)
Training -0.039
(0.124)
Explanation 0.017
(0.124)
Constant -1.518***
(0.282)		 -0.322***
(0.106)
Observations 1,068 1,068
Cond./Marg. R2 0.025/0.25 0.000/0.000
Notes: ***p < 0.01, **p < 0.05, *p < 0.1. Standard errors in parentheses.
Paper 4 research model
45

Paper 6 — Research Model

Co-Creation to Enhance Reflective Decision-Making — EJIS (submitted)
Research model: co-creation to three perceptions to three behavioral outcomes
46

Paper 6 — Variable Definitions

Co-Creation to Enhance Reflective Decision-Making — EJIS (submitted)

Co-creation
Prahalad & Ramaswamy (2004);
Vargo & Lusch (2008)
The joint process through which humans and AI systems collaboratively enhance the value of decisions by allowing users to actively shape, interpret, and refine AI recommendations.
Information Seeking
Abdel-Karim et al. (2023);
Boud et al. (1985); Schön (1984)
An objective measure of users' active interaction with the AI system to gather information for their decision.
Progressive Decision Formation
Abdel-Karim et al. (2023);
Boud et al. (1985); Schön (1984)
An objective measure for the degree to which users ultimately select alternatives that emerged through later iterations of refinement rather than defaulting to initial AI recommendations.
Decision Commitment
Abdel-Karim et al. (2023);
Boud et al. (1985); Schön (1984)
An objective measure for the likelihood that users follow through on their chosen alternative through subsequent actions.
Perceived Diversity of Alternatives
Sun et al. (2023)
Users’ subjective assessment of the diversity of alternatives recommended by the AI.
Perceived Investment in Decision Process
Norton et al. (2012)
Users’ subjective assessment of the cognitive effort and intellectual contribution they invest in shaping, interpreting, and refining decision alternatives.
Perceived Understanding of Decision
Grant et al. (2002)
Users’ subjective assessment of their understanding why particular AI recommendations emerge and what trade-offs they entail.
47

Paper 6 — Co-Creation Interface

Co-Creation to Enhance Reflective Decision-Making — EJIS (submitted)

Annotated chat interface showing recommendation, explanation, concept intervention and refined recommendation
48

Paper 6 — Study Procedure

Co-Creation to Enhance Reflective Decision-Making — EJIS (submitted)

Study procedure: introduction, main study with profile initialisation and up to eight rounds, post-survey
49

Paper 6 — Descriptive Statistics

Co-Creation to Enhance Reflective Decision-Making — EJIS (submitted)

Variable Description Black-box
(N=32)		 Explanation
(N=26)		 No Co-creation
(N=12)		 Co-creation
(N=12)
Information Seeking Number of rounds {1, 2, …, 8} 3.969 (0.361) 4.038 (0.400) 2.000 (0.589) 5.333 (0.589)
Progressive Decision Formation Position of chosen course [0, 1] 0.477 (0.075) 0.338 (0.084) 0.375 (0.123) 0.677 (0.123)
Decision Commitment Whether the user clicked on chosen course link {0, 1} 0.500 (0.085) 0.346 (0.095) 0.333 (0.140) 0.833 (0.140)
Perceived Diversity of Alternatives 7-point Likert-scale 4.115 (0.211) 4.628 (0.234) 4.056 (0.345) 5.778 (0.345)
Perceived Investment in Decision Process 7-point Likert-scale 4.552 (0.211) 4.256 (0.234) 4.944 (0.344) 5.667 (0.344)
Perceived Understanding of Decision 7-point Likert-scale 4.667 (0.207) 4.987 (0.230) 5.194 (0.338) 6.139 (0.338)
Notes: Variable means with standard errors in parentheses.
50