However, much less is known about the public\u2019s awareness of and attitudes towards the use of AI in news and journalism. In June 2024, we conducted a series of focus groups with individuals across Australia, Brazil, South Africa, and the United States, to better understand how the public defines \u201cnews,\u201d \u201cjournalism\u201d and \u201cjournalists\u201d and the role that each of them play in keeping individuals informed about important events and issues. We also explored how participants understand the role of technology, and AI in particular, in shaping their information access and environment. <\/p>\n\n\n\n
This essay \u2014 exploring how the public views artificial intelligence (AI) use in news and journalism<\/strong> \u2014 is the second in a series that highlights themes from the discussions that will inform additional research. Each focus group included participants with a mix of socio-demographics and levels of interest in keeping up with current events. More about the focus groups can be found at the end of this essay<\/mark><\/a>.<\/p>\n\n\n\n Focus group discussions of AI were centered on the use of generative AI in news, specifically. We wanted to know: How does the use of generative AI fit into participants’ understanding of news and journalism? How comfortable are participants with the use of AI in this context? Do they see it as improving or hurting their access to news and information? <\/p>\n\n\n\n AI is a complex technology that is changing rapidly, and participants had a wide range of awareness of what it is, how it works, how it can be used today, and emerging applications. Participants\u2019 knowledge \u2014 including incorrect information \u2014 played a role in shaping their opinions, both positive and negative.<\/p>\n\n\n\n We delve into these themes in detail and highlight important nuances in how the public thinks about these concepts across different contexts.<\/p>\n\n\n\n Many participants drew on their own experiences using AI to ground their perspectives about the use of the technology in news. From awareness and use to general attitudes about technology, participants\u2019 context for the discussion varied dramatically across and within groups. Several participants said that they had not used AI tools; some said that they did not feel well enough informed about the topic to weigh in on the discussion.<\/p>\n\n\n\n Of those who did have experience with AI, many were familiar with grammar tools that implement AI to improve language and word choice<\/strong>; several participants had also used tools that allow them to generate text, and described using AI to draft<\/strong> emails, letters and other documents. <\/p>\n\n\n\n \u201c…But I myself as a student, I have so much to do, I need some help, just a foundation of something that I am going to write, something to add to it.” (Brazil, from Portuguese)<\/em><\/em><\/em><\/p>\n\n\n\n \u201cI\u2019ve used it to write a couple of things that I just honestly didn\u2019t have the time to really sit and think.\u201d (Australia)<\/em><\/em><\/em><\/p>\n\n\n\n Some participants described using AI tools to help them do research and learn new information<\/strong>. In some cases, people described using AI assistants embedded in search platforms and mobile devices, as well as standalone apps like ChatGPT. Some described using these tools to replace traditional search queries, with the resulting output answering a specific question; others focused on their use of AI to get started on exploring a topic or an idea in a new way. <\/p>\n\n\n\n \u201cIf you don\u2019t\u2026quite know how to learn something, you can just chuck a prompt in there.\u201d (Australia)<\/em><\/em><\/em><\/em><\/p>\n\n\n\n \u201cIf you’re using Edge […] you will then have a chat box which is an AI, right, where you can then communicate as if you are communicating with a human. That way it helps you to narrow down what you’re asking, and it actually searches on your behalf for exactly what you are looking for.\u201d (South Africa)<\/em><\/em><\/em><\/em><\/p>\n\n\n\n Few participants said they had used software to generate images<\/strong>, and many expressed negative feelings about AI-generated or -edited images that they had encountered.<\/p>\n\n\n\n \u201cWe’ve seen or heard of images being shared online that were completely made from AI, and it’s amazing how many people don’t realize that it’s fake.\u201d (United States)<\/em><\/em><\/em><\/em><\/em><\/p>\n\n\n\n Participants across multiple groups and countries raised concerns about the impact of increasing AI usage (in general) on critical thinking skills<\/strong>, worrying that use of these tools will encourage intellectual or creative laziness among users. <\/p>\n\n\n\n \u201cIt makes us to not think actively, not think aggressively, but also not become creative in how we think. We then rely on machines to do the work.\u201d (South Africa)<\/em><\/em><\/em><\/em><\/em><\/em><\/p>\n\n\n\n These concerns were almost exclusively directed at other people\u2019s<\/em> use of AI technologies \u2014 and they were often raised in the context where learning is a key task for the end user: students, trainees, early career professionals, and journalists were all named as examples.<\/p>\n\n\n\n \u201cAs an educator, I don’t want my students using AI to write their essays. It is important for students to understand the writing process and be able to express their thoughts through writing.\u201d (United States)<\/em><\/em><\/em><\/em><\/em><\/em><\/em><\/p>\n\n\n\n \u201cI\u2019ve used ChatGPT just to ask a general question and I\u2019ve found that the answer generally is very general \u2026 The thing is, if you use that in that context, you\u2019re not actually learning because someone\u2019s giving you the answer, whereas you\u2019ve actually learned \u2026 analytical skills or research skills or how to analyse the topic, you\u2019ll actually come out better in the end.\u201d (Australia)<\/em><\/em><\/em><\/em><\/em><\/em><\/em><\/p>\n\n\n\n Many participants saw clear advantages to using AI tools to get informed on important topics and issues, using AI-assisted search tools or conversational tools like ChatGPT. Many participants described having AI summarize available and relevant information on a topic. They described it as making it \u201ceasier\u201d and \u201cfaster\u201d to learn new information. <\/p>\n\n\n\n \u201cAI will give me 10x more than what humans can give me.\u201d (South Africa)<\/em><\/em><\/em><\/p>\n\n\n\n In some cases, participants described such AI tools as a good way of getting informed on current events and issues; AI assistants embedded in search tools, in particular, played a role in providing news and information to them. <\/p>\n\n\n\n \u201cSo if you use the engines, using AI, it makes your search reliable and important. So it’s exactly what you’re getting from the radio \u2014 what you can get there and even get data and more of the in-depth information.\u201d (South Africa)<\/em><\/em><\/em><\/p>\n\n\n\n \u201cWith the AI, co-pilot, I really like it because it’s made life so much easier because what it’s done really, it’s like taking all the data from everywhere and putting it together for you […]. And from that, the AI is generating what we as humans have inputted and then it distributes to everyone, everywhere.\u201d (South Africa)<\/em><\/em><\/em><\/p>\n\n\n\n The majority of participants did not have an expectation of full accuracy when using AI tools in this way. In using AI tools to inform themselves, participants often noted that they believed additional verification steps are needed to confirm that the information they receive from generative AI tools is correct.<\/p>\n\n\n\n \u201c[The] AI is \u2026 getting different types of information throughout, sourcing that information for me, and is giving me that information. Then what I need to do, I need to then cross-check it\u2026\u201d (South Africa)<\/em> <\/p>\n\n\n\n Some participants referenced following links to sources provided by AI assistants or described evaluating the accuracy of AI-generated information based on their own knowledge. Several participants said that this is similar to the way they check information and news that they get from other non-AI sources, including radio, mobile phones and social media platforms.<\/p>\n\n\n\n Although many participants have found generative AI to be helpful for informing themselves, they expressed significantly more discomfort with journalists and news organizations using the tools to create published information. <\/p>\n\n\n\n \u201cI am skeptical of the \u2018news\u2019 generated by AI. Not opposed to it. Just approaching it with a little skepticism.\u201d (United States)<\/em><\/em><\/em><\/p>\n\n\n\n \u201cWhen you’re creating something for your personal use, it’s okay. But for professionals, journalists, it’s not acceptable.\u201d (Brazil, from Portuguese)<\/em><\/em><\/em><\/p>\n\n\n\n However, participants distinguished between many kinds of uses, and those that had the most frequent support among participants were those most similar to ones participants reported using in their own lives, suggesting that familiarity with and experience with AI technology helps people feel more comfortable<\/strong> with its use by others, an idea supported<\/mark><\/a> by existing research<\/mark><\/a>. <\/p>\n\n\n\n Some of the discomfort seemed tied to participants\u2019 expectations that journalists be highly skilled researchers and thus do their own research. This informed some participants\u2019 negative attitudes about journalists\u2019 use of AI tools:<\/p>\n\n\n\n \u201cIn my opinion, if they are using AI, then what is the need for the reporters in the first place?\u201d (United States)<\/em><\/em><\/em><\/em><\/p>\n\n\n\n \u201cSomeone who went to school for years, the journalist, someone who went to school for years now, they cannot think for themselves. They use computer software to think for them. I don’t think that’s a good thing.\u201d (South Africa)<\/em><\/em><\/em><\/em><\/p>\n\n\n\n \u201cNo, we’re talking about people who perform work in journalism. They don’t have to use AI software, because that person already has the content, the person has to write about the content, they are not creating that content, because the creation of content is considered fake news.\u201d (Brazil, from Portuguese)<\/em><\/em><\/em><\/em><\/p>\n\n\n\n However, participants often cited those same skills as reasons journalists could use AI, as long as they ensured that the information generated was accurate. <\/p>\n\n\n\n \u201cI think it’s convenient, it’s easy, but it has to be true. And you have to revise it, and you have to evaluate it. So, you don’t run the risk of being just a copy.\u201d (Brazil, from Portuguese)<\/em><\/em><\/em><\/em><\/em><\/p>\n\n\n\n \u201cI am fine with people creating content like that to save time but they need to double check because AI can’t always be reliable. Sharing is fine as well but the person who shares should verify the credibility before sharing. That goes for reporters and people working in news as well.\u201d (United States)<\/em><\/em><\/em><\/em><\/em><\/em><\/p>\n\n\n\n Some participants said that their perceptions of their local news organizations would shift to becoming less trusting of the content these organizations produced if AI was used to create it without that layer of oversight. <\/p>\n\n\n\n “If I knew that my local news organization was using AI to get their news and write their news stories, it would make me very distrusting of them.” (United States)<\/em><\/em><\/em><\/em><\/em><\/em><\/p>\n\n\n\n \u201cI think that it would somehow lose some credibility because somehow, you’re going to believe that it’s not 100% trustworthy, that you’re going to come across some misinformation there.\u201d (Brazil, from Portuguese)<\/em><\/em><\/em><\/em><\/em><\/em><\/p>\n\n\n\n Two key factors impacted participants’ general sense of when and how generative AI could be used in news, including their perception of the quality of the data available to AI, and the ability of AI models to accurately synthesize information given specific expectations for news. In the following sections, we explore these concerns.<\/p>\n\n\n\n News is about facts for most people, and whether or not AI is used, participants cared deeply about the sources of information used to produce the information that they rely on. They frequently raised questions, voiced concerns, or expressed excitement about \u201csources\u201d as they weighed both their own use of AI tools and use by journalists and news organizations.<\/p>\n\n\n\n \u201cWhat I’d like to know is, […] Is it generic or is it from a source?\u201d (South Africa)<\/em><\/em><\/em><\/p>\n\n\n\n \u201cI would like to know the sources from which the data was extracted and summarized by the AI.\u201d (United States)<\/em><\/em><\/em><\/p>\n\n\n\n \u201cDepends on where their sources come from, I suppose.\u201d (Australia)<\/em><\/em><\/em><\/p>\n\n\n\n Many participants seemed to understand that AI tools use vast amounts of data to power their models and were broadly aware that the quality and availability of data sources and the models used by AI<\/strong> have an impact on the quality of the outputs. <\/p>\n\n\n\n While participants did not always explicitly distinguish between them in conversation, there were three primary ways in which they understood data sources as playing a role in the quality of generative AI outputs. <\/p>\n\n\n\n Training data: <\/strong>Throughout the discussion, participants talking about AI\u2019s \u201csources\u201d sometimes seemed to be referring to the data used to train AI models, as in discussions of conversations with ChatGPT. As discussed above, many participants felt that this information was insufficiently reliable on its own and requires some degree of verification.<\/p>\n\n\n\n \u201cIt can be wrong because AI only uses the data it’s learned or has been fed.\u201d (Australia)<\/em><\/em><\/em><\/em><\/p>\n\n\n\n \u201cAnd so I wouldn’t rely on the journalism that comes from it [AI] because it’s information that is sourced from somewhere else and taken and put into one app.\u201d (South Africa) <\/em><\/em><\/em><\/em><\/p>\n\n\n\n Information on the Internet more broadly: <\/strong>In other cases, participants discussed the use of AI-powered search tools that help them find and summarize information from the Internet. In this case, \u201csources\u201d often seemed to refer to online content accessed by the search engine. Often, their perception of the quality of information on the Internet affected their sense of the reliability of these tools. <\/p>\n\n\n\n \u201cWhen you go to the AI and you ask for something, it gives you all of that information at once, where you don’t have to go and search on this specific site or that specific site that Google is giving you.\u201d (South Africa)<\/em><\/em><\/em><\/em><\/em><\/p>\n\n\n\n \u201cI have peace of mind because they [AI tools] are basing themselves on information that is online. So, they are real information.\u201d (Brazil, from Portuguese)<\/em><\/em><\/em><\/em><\/em><\/p>\n\n\n\n\n
![\"\"](\"https:\/\/cnti.org\/wp-content\/uploads\/2024\/08\/quotes-ai-02.png\")
<\/figure>\n<\/div>\n<\/li>\n\n\n\n![\"\"](\"https:\/\/cnti.org\/wp-content\/uploads\/2024\/08\/quotes-ai-01.png\")
<\/figure>\n<\/div>\n<\/li>\n\n\n\n![\"\"](\"https:\/\/cnti.org\/wp-content\/uploads\/2024\/08\/quotes-ai-03.png\")
<\/figure>\n<\/div>\n<\/li>\n\n\n\n![\"\"](\"https:\/\/cnti.org\/wp-content\/uploads\/2024\/08\/quotes-ai-04.png\")
<\/figure>\n<\/div>\n<\/li>\n\n\n\n![\"\"](\"https:\/\/cnti.org\/wp-content\/uploads\/2024\/08\/quotes-ai-05.png\")
<\/figure>\n<\/div>\n<\/li>\n<\/ul>Key Observations<\/h3>\n\n\n\n
\n
The Public\u2019s Personal Use<\/strong><\/h2>\n\n\n\n
Personal Use and General Attitudes About AI<\/strong><\/strong><\/h3>\n\n\n\n
Use of AI To Get Informed<\/strong><\/strong><\/h3>\n\n\n\n
Use By Journalists<\/strong><\/h2>\n\n\n\n
A Nuanced View of How Generative AI Can Be Used by Journalists<\/strong><\/strong><\/h3>\n\n\n\n
The Quality of the Data in AI Tools Matters Greatly \u2026<\/strong><\/strong><\/h3>\n\n\n\n