Can AI be a content generator? Effects of content generators and information delivery methods on the psychology of content consumers

https://doi.org/10.1016/j.tele.2020.101452Get rights and content

Highlights

  • The potentiality of using artificial intelligence (AI) technologies in news contents is addressed.

  • The experiments with a two (human vs. AI) and three (text vs. audio vs. video) were conducted.

  • Based on the results, both implications in using AI technologies are presented.

Abstract

Considering the rapidly improving technologies in artificial intelligence (AI), researchers in mass communication and journalism have paid attention to the use of AI technologies. However, there are still notable concerns about the particular use of AI technologies in this field. Put simply, can AI technologies reduce human tasks? In order to address this question, this study investigates the effects of content generators (human vs. AI) and information delivery methods (text vs. audio vs. video) on users’ perceptions of content. The results indicate that the generators and methods play a notable role in eliciting greater quality, satisfaction, and readability of the content. Based on the findings, the implications are addressed.

Introduction

Artificial intelligence (AI) has been used for media transformation (Pan, 2016, Webster and Ivanov, 2020). There have been several attempts to use AI technology for content generation in news companies (Forbes, 2018, The Washington Post, 2016). Haliograf, a robot reporter at the Washington Post, generated news contents on the Rio Olympics and the U.S. Election Day (The Washington Post, 2016). In addition, Forbes also employed its own AI publishing platform, Bertie, providing real-time trending topics (Forbes, 2018).

As news articles generated by AI have been adopted in real topics, several studies investigated how users perceive AI-generated articles (Clerwall, 2014, Wölker and Powell, 2018, Graefe et al., 2018). Clerwall (2014) investigated how readers perceived automated contents, concluding that the participants could not distinguish automated news articles from human-written articles. Moreover, they found out that the human-written articles were more pleasing to read while automated contents tended to be more informative and objective.

While considering the readers’ perception on automated contents is important along with the adoptation of AI technologies in the fields of journalism and mass communication, we need to take a broader perspective to the range of contents as well as media. People nowadays access information through various sensory media (Khan, 2017). For example, people not only watch news videos on Youtube, but also listen to the radio through Podcasts. Catching up with these technological improvements, AI can be employed to help produce various types of content.

Therefore, this study investigates the effects of content generators and information delivery methods on users’ perceptions of contents and addresses the following research question:

  • Research question: Is there a difference in the effectiveness of information delivery depending on content generators and methods?

Section snippets

AI-generated contents

As the exponential growth of computational power has provided increased support to the success of AI, high-level techniques of creating diverse contents have been developed. In the field of natural language processing, deep learning models have reached the level of creating novels, poems, or news articles (Ballardini et al., 2019, Clerwall, 2014); speech is synthesized by conversational agents embedded in portable electronic devices (Hirschberg and Manning, 2015). In addition to text and speech

Method

The experiment was a 2 (between-subjects: content generator, human vs. AI) × 3 (within-subjects: information delivery method, text vs. audio vs. video) mixed-subjects design. Sixty students (30 females and 30 males) from two large private universities in South Korea took part in the experiment. The mean age of the participants was 24.2 years (SD = 4.19), ranging from 18 to 41.

Results

A mixed multivariate analysis of variance (MANOVA) was conducted to analyze the effects of content generators and information delivery methods on the dependent variables, followed by post hoc analysis.

Discussion and conclusion

This paper investigates how individuals perceive AI-generated contents through various information delivery methods. In the case of the text contents, we assumed there would be no significant difference between content generators since AI-generated articles seemed to be high quality and well written. Consistent with the findings of prior studies (Haim and Graefe, 2017, Clerwall, 2014), the results confirmed that there is no difference in users’ perceived quality, readability, and credibility of

Acknowledgements

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. NRF-2020R1C1C1004324, NRF-2020R1F1A1048225).

References (32)

  • C. Clerwall

    Enter the robot journalist

    Journalism Practice

    (2014)
  • Forbes, 2018. Entering the next century with a new forbes experience....
  • Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., Bengio, Y., 2014....
  • A. Graefe et al.

    Readers’ perception of computer-generated news: credibility, expertise, and readability

    Journalism

    (2018)
  • M. Haim et al.

    Automated news: better than expected?

    Digital Journalism

    (2017)
  • J. Hirschberg et al.

    Advances in natural language processing

    Science

    (2015)
  • Cited by (0)

    View full text