AI partners match human partners on short virtual retail tasks, but whether they help depends on worker emotions; perceived service empathy drives how emotion translates into collaboration performance and interacts with partner type.
Adoption of AI partners in temporary tasks: exploring the effects of emotion on collaboration proficiency
In an online experiment of 861 retail employees, being paired with an AI rather than a human did not reduce collaboration proficiency for short virtual tasks, but employees' emotion moderated that relationship and service empathy mediated the link from emotion to proficiency, with partner type also altering how empathy translated into performance.
Whilst the use of artificial intelligence (AI) partners in the workplace has become more pervasive in recent years, the effects of teamwork partner type (human vs AI) on collaboration proficiency, especially for temporary tasks, remain unclear. Based on the computers are social actors (CASA) paradigm, we identify the mechanism to explain how partner type and partners' emotion influence collaboration proficiency when dealing with temporary tasks. Through an online experiment, hypotheses were examined using data collected from 861 employees working in the online retail industry. The results indicate that the type of teamwork partner does not significantly influence collaboration proficiency. However, emotion plays a significant moderating role in the relationship between partner type and collaboration proficiency. Additionally, several mediation effects are identified. Specifically, teamwork partner type moderates the effect of service empathy on collaboration proficiency whilst service empathy mediates the association between emotion and collaboration proficiency. This study is the first to reveal, from the employee's perspective, the outcomes of human–human and human–AI collaborations when dealing with temporary tasks in a virtual context. These findings can empower managers to more effectively select and pair teamwork partners, while also creating work environments that are more attuned to and supportive of emotional dynamics.
Summary
Main Finding
Using the "computers are social actors" (CASA) framework and data from an online experiment of 861 online-retail employees, the study finds that teamwork partner type (human vs AI) has no direct, significant effect on collaboration proficiency for temporary virtual tasks. Instead, emotional dynamics are central: employees' emotion moderates the partner-type → collaboration proficiency relationship, and service empathy mediates the link between emotion and collaboration proficiency. Teamwork partner type also moderates how service empathy affects collaboration proficiency.
Key Points
- Primary result: No significant main effect of partner type (human vs AI) on collaboration proficiency for temporary, virtual tasks.
- Emotion matters: Emotional state of employees significantly moderates how partner type relates to collaboration proficiency.
- Service empathy is a key mediator:
- Service empathy mediates the relationship between employee emotion and collaboration proficiency.
- Partner type moderates the effect of service empathy on collaboration proficiency.
- Theoretical anchor: Findings are interpreted through the CASA paradigm — people respond to computers as social actors, so social/affective cues (not merely whether a partner is human) shape collaboration outcomes.
- Novelty: First study to compare human–human and human–AI collaboration outcomes for temporary virtual tasks from employees’ perspective in an applied service-industry context.
Data & Methods
- Design: Online experiment conducted with employees in the online retail industry.
- Sample: 861 participants.
- Context: Virtual teamwork on temporary tasks (short-duration, task-focused collaborations).
- Analyses: Hypothesis testing via moderation and mediation analyses to investigate:
- Direct effect of partner type on collaboration proficiency.
- Moderating role of employees’ emotion on the partner type → proficiency link.
- Mediating role of service empathy between emotion and proficiency.
- Interaction (moderation) of partner type on the service empathy → proficiency relationship.
- The study emphasizes affective/social mechanisms (CASA) rather than purely capability-based explanations.
Implications for AI Economics
- Complementarity vs substitution: The lack of a main effect of AI vs human partners suggests AI teammates may substitute for humans on short, temporary tasks without clear productivity loss—conditional on emotional/empathetic factors. This nuance affects models of labor substitution and technology-induced displacement.
- Match quality and assignment algorithms: Productivity implications depend on emotional fit and perceived empathy; algorithms that assign workers to AI vs human partners should incorporate affect-related signals (e.g., current emotional state, task type, need for empathy).
- Design and valuation of AI teammates: Investing in AI features that convey empathy or supportive social cues could increase collaboration proficiency when emotion matters, raising the effective value of AI teammates beyond raw task automation.
- HR and organizational policy: Firms should consider emotional dynamics when designing hybrid teams—training, monitoring, and pairing strategies (human–human, human–AI) matter for short-term task performance. This has implications for workforce planning, task allocation, and compensation models that account for team composition effects.
- Welfare and labor-market outcomes: Because emotional and social cues shape productivity in human–AI teams, adoption decisions and regulations should consider not only efficiency gains but also worker well-being and the potential need for investments in AI social-affective design.
- Research agenda: Economic models of AI adoption should incorporate behavioral channels (emotion, perceived empathy) and heterogeneity across task temporariness and virtual settings to better predict productivity, adoption decisions, and labor demand shifts.
Assessment
Paper Typerct
Evidence Strengthmedium — Strengths: a reasonably large sample (n=861) and random assignment to partner type give credible causal evidence for the main effect of partner type. Weaknesses: outcome appears to be collaboration proficiency likely measured via self-report rather than objective productivity, emotion is apparently observed rather than randomized, mediation relies on strong assumptions, and the tasks are short, virtual, and industry-specific, limiting external validity.
Methods Rigormedium — Rigorous in using an experimental design and formal moderation/mediation analysis, but important details are missing or limiting (e.g., whether emotion was manipulated, measurement validity of collaboration proficiency and empathy, treatment realism and AI agent specification, pre-registration/balance checks, and strong assumptions needed for causal mediation).
SampleOnline experiment with 861 employees working in the online-retail/service sector who completed short-duration virtual teamwork tasks; recruitment and demographic coverage not specified in the summary (likely a convenience or industry-specific sample).
Themeshuman_ai_collab productivity
IdentificationParticipants were randomly assigned to a teamwork partner type (human vs AI), which identifies the causal effect of partner type on collaboration proficiency; moderation by employee emotion appears to use observed/measured emotions (not experimentally manipulated), and mediation via service empathy is estimated with statistical mediation (e.g., SEM), which requires additional assumptions (no unobserved confounding / sequential ignorability) for causal interpretation.
GeneralizabilityLimited to short, temporary, virtual tasks—may not generalize to long-term or co-located teamwork, Single industry (online retail) — results may not hold in other sectors or occupations, Outcome appears self-reported (collaboration proficiency) rather than objective productivity measures, Emotion likely observed (not manipulated) and may vary by culture/context—cross-cultural generalizability unclear, AI agent design/details not specified; different AI capabilities or social-affective affordances could change results
Claims (10)
| Claim | Direction | Confidence | Outcome | Details |
|---|---|---|---|---|
| Teamwork partner type (human vs AI) has no direct, significant effect on collaboration proficiency for temporary virtual tasks. Team Performance | null_result | high | collaboration proficiency |
n=861
no significant main effect of partner type on collaboration proficiency
0.6
|
| Employees' emotional state significantly moderates the relationship between partner type (human vs AI) and collaboration proficiency. Team Performance | mixed | high | collaboration proficiency |
n=861
significant moderation by employee emotion
0.6
|
| Service empathy mediates the relationship between employee emotion and collaboration proficiency. Team Performance | positive | high | collaboration proficiency |
n=861
service empathy mediates effect of emotion on collaboration proficiency
0.6
|
| Teamwork partner type moderates the effect of service empathy on collaboration proficiency (i.e., the impact of service empathy on proficiency differs by human vs AI partner). Team Performance | mixed | high | collaboration proficiency |
n=861
significant partner-type × service-empathy interaction
0.6
|
| Findings are consistent with the 'computers are social actors' (CASA) framework: people respond to computers as social actors, so social/affective cues (not just whether a partner is human) shape collaboration outcomes. Team Performance | positive | medium | collaboration outcomes (collaboration proficiency) |
n=861
empirical patterns consistent with CASA (affective cues shape collaboration outcomes)
0.36
|
| This is the first study to compare human–human and human–AI collaboration outcomes for temporary virtual tasks from employees’ perspective in an applied service-industry context. Team Performance | null_result | medium | comparative collaboration outcomes (human–human vs human–AI) in temporary virtual tasks |
n=861
author-stated novelty claim: first comparison in this applied context
0.36
|
| Data and methods: the study used an online experiment with 861 online-retail employees performing short-duration, virtual, task-focused collaborations; analyses focused on direct effects, moderation (emotion and partner type), mediation (service empathy), and moderated-mediation. Other | null_result | high | NA (methodological claim about study design and analyses) |
n=861
methodological description of experiment and analyses
0.6
|
| Implication for substitution: Because there was no main effect of partner type on collaboration proficiency, AI teammates may substitute for humans on short, temporary tasks without clear productivity loss—conditional on emotional and empathetic factors. Team Performance | mixed | medium | productivity / collaboration proficiency |
n=861
no main effect suggests potential substitution without productivity loss conditional on emotional/empathetic factors
0.36
|
| Design implication: Investing in AI features that convey empathy or supportive social cues could increase collaboration proficiency when emotion matters. Team Performance | positive | medium | collaboration proficiency |
n=861
investing in empathy-signaling AI features could increase collaboration proficiency
0.36
|
| HR and organizational implication: Firms should consider emotional dynamics when designing hybrid teams; training, monitoring, and pairing strategies (human–human, human–AI) matter for short-term task performance. Team Performance | positive | medium | short-term task performance / collaboration proficiency |
n=861
managerial recommendation based on experimental findings
0.36
|