How effective is AI augmentation in human–AI collaboration? Evidence from a field experiment

Information Technology and People, Journal Year: 2024, Volume and Issue: 37(7), P. 2357 - 2389

Published: Nov. 6, 2024

Purpose Companies increasingly leverage artificial intelligence (AI) to enhance human performance, particularly in e-commerce. However, the effectiveness of AI augmentation remains controversial. This study investigates whether, how and why enhances agents’ sales through a randomized field experiment. Design/methodology/approach conducts two-by-two factorial experiment ( N = 1,090) investigate effects on sales. The compares outcomes handled solely by agents with those augmented AI, while also examining moderating effect experience levels underlying mechanisms behind responses. Findings results reveal that leads significant 5.46% increase Notably, impact varies based levels, inexperienced benefiting nearly six times more than their experienced counterparts. Mediation analysis shows improves response timeliness, accuracy sentiment, thereby boosting Originality/value highlights role human–AI collaboration, demonstrates varying impacts offers insights for organizations regulate agent responses drive

Language: Английский

---

When do service employees smile? Response‐dependent emotion regulation in emotional labor

Journal of Organizational Behavior, Journal Year: 2021, Volume and Issue: 42(9), P. 1202 - 1227

Published: Sept. 6, 2021

Summary We advance the theoretical and practical understanding of affect in service interactions by conceptualizing employees customers as concurrent participants same interaction. analyzed employees' emotional labor requirements, which comprise both well‐recognized requirement to display positive (i.e., acting is response independent) less‐recognized attend customers' affective displays dependent). found support for our hypotheses across two studies, compare within interactions. In Study 1, we examined field data comprising 1 320 392 customer employee messages from 164 899 real‐life chat‐based used automated sentiment analysis identify negative affect. 2, simulated examine discrete emotions. Using different methodologies, Studies 2 that differ their response‐dependent behaviors. also demonstrated behavior improves outcomes. Our analyses enrich labor, contribute theory social considering yet differing behaviors partners interaction, suggest new exciting methods Organizational Behavior research.

Language: Английский

---

‘I do not know’: an examination of reactions to virtual agents that fail to answer the user’s questions

The International Review of Retail Distribution and Consumer Research, Journal Year: 2023, Volume and Issue: 34(2), P. 228 - 250

Published: Aug. 31, 2023

Virtual agents (VAs) used by retailers for online contacts with customers are becoming increasingly common. So far, however, many of them display relatively poor performance in conversations users – and this is expected to continue still some time. The present study examines one aspect between VAs humans, namely what happens when a VA openly discloses its knowledge gaps versus it makes attempt conceal setting which cannot answer user questions. A between-subjects experiment manipulated VA, perceived service quality as the main dependent variable, shows that high level ability questions boosts quality. also offers explanations outcome terms mediating variables (perceived competence, openness disclose own limits, usefulness, learning-related benefits).

Language: Английский

---

Developing a multi-variate prediction model for COVID-19 from crowd-sourced respiratory voice data

Published: Aug. 11, 2024

Aim: COVID-19 has affected more than 223 countries worldwide and in the post-COVID era, there is a pressing need for non-invasive, low-cost, highly scalable solutions to detect COVID-19. This study focuses on analysis of voice features machine learning models automatic detection Methods: We develop deep model identify from recording data. The novelty this work development identification only recordings. use Cambridge Sound database which contains 893 speech samples, crowd-sourced 4,352 participants via Sounds app. Voice including Mel-spectrograms Mel-frequency cepstral coefficients (MFCC) convolutional neural network (CNN) Encoder are extracted. Based data, we classification cases. These include long short-term memory (LSTM), CNN Hidden-Unit BERT (HuBERT). Results: compare their predictive power baseline models. HuBERT achieves highest accuracy 86% AUC 0.93. Conclusions: results achieved with proposed suggest promising diagnosis recordings when compared obtained state-of-the-art.

Language: Английский

---

Optimizing MFCC parameters for the automatic detection of respiratory diseases

Applied Acoustics, Journal Year: 2024, Volume and Issue: 228, P. 110299 - 110299

Published: Sept. 20, 2024

Language: Английский

---

The psychological and ethological antecedents of human consent to techno-empowerment of autonomous office assistants

AI & Society, Journal Year: 2022, Volume and Issue: 38(2), P. 647 - 663

Published: July 15, 2022

Language: Английский

---

Servitization and Digitalization as “Siamese Twins”: Concepts and Research Priorities

Springer eBooks, Journal Year: 2022, Volume and Issue: unknown, P. 967 - 989

Published: Jan. 1, 2022

Language: Английский

---

Artificial Intelligence and Extended Reality in Luxury Fashion Retail: Analysis and Reflection

Springer series on cultural computing, Journal Year: 2023, Volume and Issue: unknown, P. 323 - 348

Published: Jan. 1, 2023

Language: Английский

---

Comparing Neural Networks for Speech Emotion Recognition in Customer Service Interactions

2022 International Joint Conference on Neural Networks (IJCNN), Journal Year: 2022, Volume and Issue: unknown, P. 1 - 8

Published: July 18, 2022

Automatic speech emotion recognition (SER) may assist call center service employees in deciphering and regulating customer emotions. In order to contribute a successful augmentation of with AI, the main goal this study is identify effective machine learning approaches classify discrete basic emotions conversations. A comparison presented performance different neural network architectures on features extracted from interactions naturalistic setting. Baseline classifiers, including zerorule classifier, random frequency nonsequential multi-class classifiers are compared architectures. multi-layer perceptron (MLP), one-dimensional convolutional (CNN), translation (NMT) outperform baseline suggesting pattern data relating labels. While model attention attains highest f1-score, no significant difference among networks detected. Results therefore support use multi-label as simplest model.

Language: Английский

---

Privacy-Preserving and Scalable Affect Detection in Online Synchronous Learning

Lecture notes in computer science, Journal Year: 2022, Volume and Issue: unknown, P. 45 - 58

Published: Jan. 1, 2022

Language: Английский

---