From Efficiency to Empathy: How the Integration of AI Affects Physicians’ Emotional Labor

Last registered on December 09, 2025

View Trial History

Pre-Trial

Trial Information

General Information

Title

From Efficiency to Empathy: How the Integration of AI Affects Physicians’ Emotional Labor

RCT ID

AEARCTR-0017398

Initial registration date

December 04, 2025

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published

December 09, 2025, 7:35 AM EST

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

Country

China

Region

Anhui Province

Primary Investigator

Name

Zhensheng Li

Affiliation

Anhui University of Finance and Economics

Contact Primary Investigator

Other Primary Investigator(s)

Additional Trial Information

Status

In development

Start date

2025-12-10

End date

2026-01-15

Keywords

Health

Additional Keywords

JEL code(s)

Secondary IDs

Prior work

This trial does not extend or rely on any prior RCTs.

Abstract

This study focuses on physician-patient interactions within internet hospitals. It aims to develop an AI system based on multi-dimensional emotion recognition (e.g., keywords, input frequency, pause duration) to assist physicians in real-time perception of patient emotional states. We collaborated with a major Chinese hospital, designing 20 typical interaction scenarios within internet hospitals and employing a randomized controlled trial to evaluate the impact of different AI intervention styles on physicians' emotional labor.

External Link(s)

Registration Citation

Citation

Li, Zhensheng. 2025. "From Efficiency to Empathy: How the Integration of AI Affects Physicians’ Emotional Labor." AEA RCT Registry. December 09. https://doi.org/10.1257/rct.17398-1.0

Sponsors & Partners

Experimental Details

Interventions

Intervention(s)

Intervention (Hidden)

Intervention Start Date

2025-12-20

Intervention End Date

2026-01-10

Primary Outcomes

Primary Outcomes (end points)

The emotional labor situation of doctors after AI intervenes in doctor-patient interaction.

Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)

Secondary Outcomes (explanation)

Experimental Design

The experimental procedure can be treated as a 2×2 between-subjects factorial design, systematically varying two forms of AI assistance: Whether the AI provided direct suggestions or not, and whether it supported cognitive construction or not. This resulted in four experimental conditions, allowing us to isolate the individual and combined effects of these AI intervention types. To ensure valid intervention exposure across groups, physicians were randomly assigned to one of the four conditions. All participants engaged with the same set of standardized patient scenarios, ensuring comparability in task context and emotional demands. This design aims to systematically reveal which AI intervention method can more effectively shift physicians’ emotional labor patterns from depleting surface acting to sustainable deep acting, thereby ensuring the quality of empathetic care while alleviating their emotional burden.

Experimental Design Details

In the direct suggestion conditions, the AI system proactively provided ready-to-use empathetic responses during simulated consultations. In the cognitive construction conditions, the AI instead delivered scenario frameworks for understanding patient perspectives, without supplying explicit wording. Under the control condition (no direct suggestion × no cognitive construction), physicians conducted conversations without any form of AI support.

Randomization Method

The subjects were stratified and grouped by computer based on their characteristics.

Randomization Unit

Individual

Was the treatment clustered?

Yes

Experiment Characteristics

Sample size: planned number of clusters

The subjects will be randomly divided into 4 groups to ensure that their basic characteristics have no significant impact on the main results.

Sample size: planned number of observations

150-200 individuals.

Sample size (or number of clusters) by treatment arms

30-50 individuals for control, 30-50 individuals enabled direct-suggestion AI, 30-50 individuals enabled cognitive-construction AI, 30-50 individuals for both treatment.

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Supporting Documents and Materials

IRB