Italy is not exactly at the forefront of AI usage and development, although several researchers, professors and experts are well known in the field, in some cases since decades, are Italian – among them, just to choose two, professor Nello Cristianini, author of two important books about generative artificial intelligence and professor Giuseppe Attardi: see his lecture Large Language Models are All You Need?.

The recently published document “Italian Strategy for Artificial Intelligence 2024-2026” recognises that AI can have a transformative impact on the country’s socio-economic system, benefitting from the rapid advancements and great opportunities offered by generative AI. To harness AI’s benefits, however, there is a significant need for highly skilled professionals capable of developing and managing AI algorithms and systems. Currently, the availability of such skills in Italy is far below demand, which significantly slows down the adoption of these solutions.

The journalism sector in Italy is aligned with this situation: it’s still grappling with a digital transition that has never been fully realised, with business models that often look to the past, and in some cases, an approach that is not yet digital-first, for several reasons.

Thus, it’s crucial to focus on training those who will educate and inform the public, starting from journalists and publishers.

Survey insights

That’s the reason why, in recent months, in collaboration with the Lombardy branch of the Italian Order of Journalists and working on an event in partnership with the U.S. Consulate in Milan, and Slow News, the magazine I direct, I have personally conducted a survey that gathered over 600 responses – and counting – from media professionals to monitor their perceptions, educational needs, fears, and expectations on the artificial intelligence field.

Despite high interest, with over 85% of respondents recognising the need for further education on AI applications, only about 36% have actually attended AI training events.

This stark contrast between high demand for AI training and low participation in such events highlights a critical gap in the availability or awareness of comprehensive AI education tailored for media professionals. This also underscores the necessity for more structured, continuous educational programs that extend beyond introductory sessions.

Among these, 50% participated in one-time free online courses, suggesting that while there is accessibility in terms of cost, there is a lack of control on the quality of these training, a lack of ongoing projects or advanced training opportunities.

Only in 14.5% of cases the training was paid by the publisher. In 34.5% of cases, people are training themselves. A good role is played by the free courses provided by the Order of Journalists: in Italy, there is a legal requirement for continuous training which involves the need to follow at least 20 hours of professional training in a year, which is important but not sufficient.

Specific groups like photojournalists and other specialised journalists have shown significant interest in practical AI applications, suggesting that targeted workshops and hands-on training could be exceptionally beneficial.

Only a minority (11%) of the environments reported having a dedicated AI unit, indicating that while individual interest in AI is substantial, institutional adoption and integration lag behind. This could be an area for development, where organisations not only adopt AI technology but also invest in the necessary training and structural changes to support its effective use.

It’s noteworthy that many journalistic entities in the world, including Italy as seen from the robots.txt files of major newspapers like La Repubblica, Corriere della Sera, and Il Sole 24 Ore, have blocked AI bots – check the robots.txt of each website, following the links: it’s a way to know if your newspaper is blocking them or not. Despite this, more than 62% of respondents are unaware if their newspapers have blocked AI bots or not.

An excerpt from a robots.txt file from an Italian newspaper. Lines of code with “disallow” and the names of various artificial intelligence bots (GPTbot, anthropic-ai, …) indicate that those bots are blocked, i.e. that the newspaper does not want its texts to be scanned and possibly included in training or fine tuning of a large language model

In the vast majority of cases, Italian journalists are not involved in these kinds of decisions.

Moreover, 42% have never even discussed an AI usage policy, 47% have had just informal discussions, and only a small fraction (5%) have engaged in formal meetings or have a clear, published and transparent policy for the readers (1,8%).

Deficiencies and interests

In response to the question about their own deficiencies on the topic (with the possibility of choosing multiple answers), 43% say they do not have the knowldge to choose the tools to use, 34.5% do not know how to use them, 23% lack basic knowledge, 23.6% are unaware of the social implications, and 24.4% do not understand the political implications. Additionally, 3% admit they do not even understand why artificial intelligence should be of interest to them.

72.9% are interested in artificial intelligences that generate text, 53.4% in those that generate images, 49% in data analysis, and 44% in the possibility of using AI for fact-checking.

It is clear that the immense hype surrounding ChatGPT, Gemini, and other chatbots has profoundly influenced the way Italian journalists envision using artificial intelligence. This often overshadows other uses that are far more interesting for journalism, such as consulting databases for information extraction.

Positive aspects and fears

Regarding the benefits of AI, while a minor 6.3% couldn’t identify any positive aspects, over 93% acknowledged various benefits (people were allowed to choose multiple answers). Here the results:

Time Savings (63.5%): many journalists appreciate AI’s ability to automate routine tasks, such as data sorting and basic reporting, which saves time and allows journalists to focus on more in-depth investigative work.
Increased efficiency in content production (49.3%): AI tools can streamline the content creation process, from research to writing drafts, thereby enhancing productivity across editorial teams.
Faster information retrieval (40.6%): AI can quickly analyse large datasets, helping journalists gather facts and figures in real-time, which is crucial in breaking news situations.
Quicker summarisations (33.5%): AI’s ability to summarise lengthy reports or articles helps journalists distil essential information, making it easier to update or repurpose content for different formats.
Creation of new content formats (29.2%): AI is paving the way for innovative content formats, such as automated video scripts or interactive articles, which can attract diverse audiences.
Cost-effective junior assistant (27.7%): AI tools act as low-cost “junior assistants” that handle tedious tasks, reducing the need for larger staff while maintaining content output.
Enhanced brainstorming (24.92%): AI can help generate creative ideas for stories or headlines, sparking creativity in editorial discussions.
Faster social media content (23.10%): with AI, media outlets can quickly adapt stories for various social media platforms, enhancing their reach and engagement.
Improved accuracy and source verification (21.78%): AI can assist in fact-checking and verifying sources more efficiently than manual processes, which is vital for maintaining journalistic integrity.
Personalised content (21.29%): AI algorithms can help tailor content to individual preferences, providing a more personalised experience for the audience.

Conversely, the biggest concerns include the creation of fake and deep fake content, ethical and privacy issues, proliferation of low-value content, lack of human oversight, and job losses.

Fake and deepfake content: The potential for AI to create realistic but entirely fabricated images, videos, or stories is a profound concern, as it could lead to misinformation and damage public trust.
Ethical and privacy issues: There are fears about how AI might compromise journalistic ethics, particularly regarding privacy violations and the potential misuse of personal data.
Proliferation of low-value content: There’s a worry that AI might flood the market with superficial or low-quality content, overshadowing thorough, researched journalism.
Lack of human oversight: The autonomy of AI systems can lead to situations where human judgement is sidelined, which is risky in a field as sensitive as journalism.
Job losses: There is an underlying fear that AI could replace human jobs, particularly in areas like basic reporting and content generation, leading to reduced employment opportunities in the industry.

It’s interesting to note that practically all the positive aspects and expectations regarding the possibility of using these technologies pertain to very practical issues. On the other hand, all the negative aspects and fears regard theoretical issues, which is actually the first part to work on: for the very nature of AIs and LLMs it’s hard to understand how to use them if you don’t know – at least at a very superficial level – how they work and why they give you certain results, why they hallucinate and therefore how to properly find a way to apply these powerful tools to the daily workflow.

Open answers

As often happens, the open-ended responses (left by the 17% of the respondents) in the “Artificially Informed” survey provide a rich qualitative insight into the nuanced views and specific needs of media professionals regarding AI.

1. Practical application demands

There is a clear request for workshops that not only explain the theory behind AI technologies but also demonstrate their application in real journalistic tasks. For example, a workshop on how AI can be used to automate the routine aspects of journalistic writing, thereby freeing up time for creative and investigative endeavours.

2. Ethical and social implications:

A significant number of responses addressed the need for discussions on the ethical use of AI in journalism. Respondents expressed concerns about bias in AI algorithms and the potential for AI to be used in ways that could undermine public trust in the media. There is a call for training that covers how to identify and mitigate bias in AI applications, ensuring that AI tools are used in a manner that aligns with ethical journalistic practices.

3. Policy and transparency

Responses indicated a desire for more clarity on policies regarding AI usage within organisations. Many participants expressed uncertainty about the existing policies or noted a lack of discussion about how AI should be ethically integrated. This suggests a gap in communication within organisations about AI, highlighting the need for clear guidelines and open discussions on how AI technologies are implemented and governed.

4. Advanced AI training

While the vast majority of participants are still seeking basic understanding of AI, others are looking for more advanced training, reflecting a diverse range of knowledge and comfort with AI within the industry. Advanced topics of interest include the use of AI for complex data analysis, machine learning models for predictive analytics, and the creation of interactive AI-driven digital content.

Among the more unique responses were concerns about AI replacing human jobs, the need for AI to support rather than replace human creativity, suggestions for using AI to create new forms of interactive storytelling, the development of AI tools that could help journalists analyse large volumes of data to uncover stories that would be impossible to detect without technological assistance.

5. Requests for regular updates and continuous learning

A recurring theme is the need for ongoing education about AI, reflecting the fast-evolving nature of AI technologies. Respondents requested regular updates on the latest AI developments and continuous learning opportunities to keep pace with technological advancements.

Listen and train

These results could be very important to understand the journalistic ecosystem reaction to AIs, but also to proper design training. In an audience listening process based on this data extraction, that’s exactly what I’m doing, for example, hosting a first event (Artificially Informed) answering the most frequent questions and trying to feed this appetite for information about the AIs among the newsrooms and freelance journalists in Italy.

Hopefully, spreading these insights, practice and sharing experiences could be an accelerator for the diffusion of awareness and skills.

Source of the cover photo: Presentation of Artificially Informed survey’s result in Fondazione Giangiacomo Feltrinelli, Milano, picture by U.S. Consolate, Facebook

transparency

Latest in transparency