Not My Truce: Personality Differences in AI-Mediated Workplace Negotiation

Veda Duddu; Jash Rajesh Parekh; Andy Mao; Hanyi Min; Ziang Xiao; Vedant Das Swain; Koustuv Saha

cs.HCApr 1, 2026

Not My Truce: Personality Differences in AI-Mediated Workplace Negotiation

Authors: Veda Duddu, Jash Rajesh Parekh, Andy Mao, Hanyi Min, Ziang Xiao, Vedant Das Swain, Koustuv Saha

Organizations: University of Illinois Urbana-Champaign, Urbana, IL, USA · Johns Hopkins University, Baltimore, MD, USA · New York University, New York, NY, USA

Abstract

AI-driven conversational coaching is increasingly used to support workplace negotiation, yet prior work assumes uniform effectiveness across users. We challenge this assumption by examining how individual differences, particularly personality traits, moderate coaching outcomes. We conducted a between-subjects experiment (N=267) comparing theory-driven AI (Trucey), general-purpose AI (Control-AI), and a traditional negotiation handbook (Control-NoAI). Participants were clustered into three profiles -- resilient, overcontrolled, and undercontrolled -- based on the Big-Five personality traits and ARC typology. Resilient workers achieved broad psychological gains primarily from the handbook, overcontrolled workers showed outcome-specific improvements with theory-driven AI, and undercontrolled workers exhibited minimal effects despite engaging with the frameworks. These patterns suggest personality as a predictor of readiness beyond stage-based tailoring: vulnerable users benefit from targeted rather than comprehensive interventions. The study advances understanding of personality-determined intervention prerequisites and highlights design implications for adaptive AI coaching systems that align support intensity with individual readiness, rather than assuming universal effectiveness.

Figures & tables

Sex
	Overall	Trucey	Control-AI	Control-NoAI
Male	141	71	33	37
Female	121	61	32	28
Other/Prefer Not to Say	5	2	1	2
Race/Ethnicity
White	181	94	38	49

Table 1. Demographic distribution of

N

=267 participants in the experimental study. table This table provides a detailed demographic breakdown of the 267 participants in the experimental study. It categorizes the participants based on several key characteristics, including Sex, Race/Ethnicity, Educational Level, Employment Type, Work Domain, Compensation Type, and Work Experience. The data is presented for the overall study sample as well as for each of the three intervention conditions: Trucey, ChatGPT, and the traditional Handbook. The table shows the distribution of participants across these demographic and professional categories, ensuring the study's sample reflected a diverse set of backgrounds.

Figure 1. Example screenshots of Trucey through various stages of user interactions. table This figure shows a series of screenshots from the Trucey prototype, demonstrating the sequential workflow a user experiences during a negotiation coaching session. The first screenshot, labeled "Stage 1: Understanding Your Situation," give the user their pre-assigned contextual details of a negotiation scenario, such as the problem to discuss, the person they're talking to, and their relationship with that person. The second screenshot, "Stage 2: Analysing Leadership Style," shows a structured assessment where the user is asked to describe their boss’s nature and leadership style. This assessment, based on Big-Five personality traits, enables the system to tailor its strategic advice and simulation behavior to the specific supervisor's profile. The third screenshot, "Stage 3: Conversation Practice," is where the core interaction occurs. It is divided into two phases: Phase 1: Getting Advice and Phase 2: Rehearsal. In the "Getting Advice" phase, Trucey provides strategic coaching through incremental dialogue, introducing negotiation concepts gradually to prevent user overwhelm. The "Rehearsal" phase is an interactive practice session where the AI embodies the supervisor's personality profile to create a realistic negotiation experience. This phase also includes an "Iterative Feedback Integration" step, which is shown at the bottom of the last screenshot. This feature allows users to provide direct feedback to the chatbot on its realism and tone. This feedback then helps to recalibrate the system for subsequent responses. The goal of this step is to make the coaching system more situationally responsive and to maintain a shared understanding between the user and the AI. The system is technically implemented using GPT-4.1 with a Python Streamlit interface and a SQL database to facilitate these personalized interactions.

Panel of Figure 2 — Figure 2. Figures describing the clustering and clusters: (a) Elbow method (b) Silhouette scores for determining the optimal number of personality clusters. (c) PCA-based distribution of the clusters across two PCA dimensions, and (d) Heatmap showing the distribution of the personality traits across the three clusters.

Cluster	N	Extraversion	Agreeableness	Conscientiousness	Neuroticism	Openness
$\mathtt{C_{0}}$	74	-0.52	-1.04	-0.96	+0.94	+0.06
$\mathtt{C_{1}}$	110	+0.61	+0.58	+0.61	-0.66	+0.56
$\mathtt{C_{2}}$	83	-0.35	+0.16	+0.04	+0.03	-0.80

Table 2. Cluster centroids (

z

-scored) for Big Five personality traits (N=267).

$\Delta$ Metric	$\mathtt{C_{0}}$	$\mathtt{C_{1}}$	$\mathtt{C_{2}}$	H-stat.
Organizational Self-Efficacy (OSS)	-0.041	0.012	0.011	0.17
Psychological Empowerment (PEU)	0.152	0.002	0.166	4.79
PEU: Competence	0.099	0.027	0.036	0.60
PEU: Meaning	0.223	0.018	0.199	1.47
PEU: Impact	0.189	0.015	0.258	5.09
PEU: Self-Determination	0.095	-0.055	0.169	2.81

Table 3. Pooled cluster comparison across all conditions. Kruskal-Wallis tests examining whether personality clusters differed on outcome and perception measures when pooled across all conditions.

p

-values reported after Bonferroni correction. ***

p<.001

$\Delta$ Metric	Cluster	Trucey	Control-AI	Control-NoAI	H-stat.
Organizational Self-Efficacy (OSS)	$\mathtt{C_{0}}$	-0.090	-0.125	0.111	0.96
	$\mathtt{C_{1}}$	0.047	0.037	-0.090	0.96
	$\mathtt{C_{2}}$	0.200	-0.266	-0.050	8.68
Psychological Empowerment (PEU)	$\mathtt{C_{0}}$	0.087	0.174	0.248	1.93
	$\mathtt{C_{1}}$	-0.086	-0.088	0.285	8.26
	$\mathtt{C_{2}}$	0.099	0.054	0.427	3.71

Table 4. Within-cluster condition comparisons. Kruskal-Wallis

H

tests examining whether conditions produced different outcomes within each personality cluster. Significant omnibus effects are bolded; *

p

< 0.05, **

p

< 0.01, ***

p

< 0.001

Measure	Cluster	Trucey	Control-AI	Cohen’s d	t-test
Verbosity
Verbosity: Words per sentence	$\mathtt{C_{0}}$	19.90	20.14	-0.02	-0.14
	$\mathtt{C_{1}}$	19.30	22.83	-0.34	-2.53
	$\mathtt{C_{2}}$	19.53	18.53	0.13	0.84
Verbosity: Words per response	$\mathtt{C_{0}}$	38.38	45.24	-0.30	-1.85
	$\mathtt{C_{1}}$	53.02	63.10	-0.34	-2.48

Table 6. Linguistic characteristics of negotiation outputs: Trucey vs. Control within each cluster, with Cohen’s

d

effect sizes and

t

-tests reported. Significant effects are bolded. Measures are grouped by significance status. *

p

< 0.05, **

p

< 0.01, ***

p

< 0.001.

Measure	Cluster	Trucey	Control-AI	Cohen’s d	t-test
Strategy Development
Strategy Development: Basic	$\mathtt{C_{0}}$	0.17	0.18	-0.08	-0.52
	$\mathtt{C_{1}}$	0.18	0.17	0.14	1.01
	$\mathtt{C_{2}}$	0.19	0.18	0.05	0.30
Strategy Development: Structured	$\mathtt{C_{0}}$	0.26	0.30	-0.27	-1.68
	$\mathtt{C_{1}}$	0.26	0.31	-0.39	-2.84	**

Table 7. Brett negotiation framework element adoption: Trucey vs. Control within each cluster, with Cohen’s

d

effect sizes and

t

-tests reported. *

p

< 0.05, **

p

< 0.01, ***

p

< 0.001.

Human-AI Interaction

Human-AI interaction studies how people and AI systems communicate, collaborate, make decisions, and shape one another over time.

See all 329 papers

AI-Mediated Negotiation: Design Reflections and Lessons
Jun 20, 2026Veda Duddu, Jash Rajesh Parekh, Andy Mao +4Human-AI Interaction Human-Centered AI
University of Illinois Urbana Champaign, USA · University of Illinois Urbana-Champaign, USA · Johns Hopkins University, USA +1
Personality Engineering with AI Agents: A New Methodology for Negotiation Research
May 19, 2026Michelle A. Vaccaro, Jared R. CurhanAutomated Negotiation
MIT Institute for Data, Systems, and Society · MIT Sloan School of Management
Imperfectly Cooperative Human-AI Interactions: Comparing the Impacts of Human and AI Attributes in Simulated and User Studies
Apr 17, 2026Myke C. Cohen, Mingqian Zheng, Neel Bhandari +6Human-AI Interaction
Aptima, Inc. · Arizona State University · Carnegie Mellon University

Jun 20, 2026J/K move · Enter open · S save

AI-Mediated Negotiation: Design Reflections and Lessons

Veda Duddu, Jash Rajesh Parekh, Andy Mao, Hanyi Min+3

University of Illinois Urbana Champaign, USA · University of Illinois Urbana-Champaign, USA · Johns Hopkins University, USA+1

Human-AI Interaction Human-Centered AI Conversational Agents Human-AI Collaboration

Open paper AskPDF arXiv

Conversational AI promises a new kind of preparation for high-stakes workplace negotiations -- personalized, interactive, and capable of simulating realistic resistance. That promise is intuitive. We built Trucey, a theory-driven coaching system, to test it. The system encoded four assumptions: that articulation supports clarification, that personalization builds strategic competence, that chunked delivery reduces cognitive load, and that structured scaffolding removes metacognitive burden. A pre-registered experiment (N=267) and interviews (N=15) complicated each of them. Notably, the static handbook we included as a passive control outperformed both AI conditions on empowerment and usability. We reflect on why: each assumption encoded a specific model of how preparation unfolds, and the findings revealed that conversational AI imposes a linear execution model on a task that is fundamentally recursive. We identify an unexamined scope condition on established HAI design guidelines and close with a sequencing principle -- map before path, path before simulation -- for future AI coaching design.

Abstract

Human-AI Interaction

Explore similar work

AI-Mediated Negotiation: Design Reflections and Lessons

Personality Engineering with AI Agents: A New Methodology for Negotiation Research

Imperfectly Cooperative Human-AI Interactions: Comparing the Impacts of Human and AI Attributes in Simulated and User Studies