AI: risks worry parents, but can be mitigated – 7/5/2023 – Tech

Accompanying their children in their digital literacy is a long-standing concern of Natália Ruschel, 39. Attention has taken on new contours amid the popularization of AI (artificial intelligence).

“As my children are still young, 9 and 3 years old, I haven’t introduced ChatGPT to them yet, because I don’t see the need. However, I’m preparing for it.”

Ruschel works with intellectual property at Mdic (Ministry of Development, Industry, Commerce and Services), and frequently comes across the subject of artificial intelligence. She is aware of its possibilities and risks —many of which are already familiar to the online presence.

Researcher Sonia Livingstone, linked to Unicef ​​(United Nations Children’s Fund), classifies the risks of children’s online presence into four major areas, which she calls the 4Cs:

• content: engagement or exposure to content that is detrimental to your development;

• contact: possibility of being the target of harmful contact from others;

• conduct: when they witness, participate in or are victims of harmful conduct, such as violence, racism and LGBTophobia

• It is contract: terms of use, policies and privacy without transparency, and which can lead to exposure of children and adolescents.

Some AI risks are more discussed, such as recommendations for inappropriate or violent content on social media; but the debate is still in its infancy for generative artificial intelligences, which create texts and images from already existing information.

ChatGPT, released in November 2022, is one of the most talked about examples. The tool quickly responds to user queries, giving the impression of reliability — but with frequent errors and inaccuracies, which contribute to misinformation.

In recent months, Ruschel has been testing the tool, looking forward to the role it will play in his children’s future. She intends to accompany them in its use, encouraging the critical analysis of the results, but recognizes that it will not be an easy task.

“ChatGPT already does text analysis, interpretation or summary of information, it is not the result of a simple search. It will be a great challenge to know what is false, true and biased without having prior knowledge about the researched subject.”

Business model favors misinformation

Gabriela Mora, teen development and participation officer at Unicef ​​(United Nations Children’s Fund), points out that the platforms’ business model benefits from the attention generated by extremist and often false content.

AI’s lack of transparency is also a challenge, especially when considering terms of use and associated contracts. Mora says that, in addition to the family, companies and public management also have the responsibility to consider the rights of children and adolescents in their work.

“Today it is very uneven. The terms of use and privacy policies end up requiring the responsibility of the user alone.”

Responsible action by companies must begin with the programming of tools, which bear the biases of their programmers and the data that feed artificial intelligence. Such biases, associated with the business model of digital platforms, may favor the reproduction of prejudice and violence.

At this point, it is essential that reference adults serve as support. Care and use limitations must be associated with guidelines, as explained by Mora.

“You can only learn using the internet. So it is also important, as the child and adolescent matures, to warn about these risks and support them so that they can build their resilience strategy, to face issues of violence and that they can also find your care repertoire.”

AI information enables beneficial uses

Guiding children through this scenario is not easy. For Mora, the challenge starts right away in the generational aspect: “Whoever is a child and teenager today has not known the world without a smartphone, and this brings a familiarity with technology that is very different from parents and guardians, especially if we think about the socioeconomic inequality of who had access recently.”

However, familiarity does not mean maturity to deal with technology. “It requires a critical sense, a capacity for understanding. Thinking of the internet as a large public square, no one lets their children out there without proper recommendations, without due fear and care, and everyone is watching,” compares the specialist. .

In this context, it is worth learning new skills and staying informed about the risks and functioning of artificial intelligence. This can help to understand the suitability of different tools for different phases of development, making it possible to reduce their risks and take advantage of their potential.

This is the case of programmer Luciano Santa Brígida, 35. User and researcher of artificial intelligence, his biggest concern is with the content automatically recommended for his six-year-old daughter.

“I restrict what content my daughter can or cannot consume through tools provided by the platform itself, creating a child profile with content filtered according to her age. I also always review the history weekly to assess whether the algorithm may have recommended something inappropriate that I need to block or talk to her about it”, she says.

Generative tools are also part of Luciano and his daughter’s routine. “I use her drawings as a basis for image generators, or she asks me to create characters she dreams of in her games, and I take the opportunity to dialogue with her about art, styles, history and related content”, says Luciano.

However, Luciano still does not allow generative text AIs such as ChatGPT to be accessed by his daughter. “I’m encouraging her to write in her own words first,” she says.

Dialogue is irreplaceable in digital education

“I think that nothing replaces the dialogue about how artificial intelligence is present in the child’s life, at school, in games, what kind of personal information they are making available”, says Mora.

The specialist says that parental control and screen time restriction applications can be useful, but that parental monitoring is indispensable – especially considering the commercial format of the platforms, designed to keep us connected continuously.

“The child does not have the ability to self-regulate in that sense, so there is work there that cannot be outsourced. It is boring work, because the platforms will have a much greater appeal for the child to stay connected, and then you really have to there should be a limit, a negotiation”, he points out.

Still, it can be a challenge to balance guidance to adolescents with respect for their autonomy and privacy. “There’s an age that doesn’t have to have a cell phone, doesn’t have to have a screen, but as teenagers grow and mature and improve their learning process, even in virtual environments, it is the role of this reference adult in life to talk about what happens online and offline.”

Mora suggests running away from prohibitions and moral lessons. “A list of prohibitions is easily circumvented by teenagers. Instead, build with your children a notion of self-care, care for other people, alerting them to legal and emotional consequences and responsibilities, issues of violence and the support network.”

Use in education must be supervised, not prohibited

Researcher Tainá Aguiar Junquilho, from the ITS (Rio Institute of Technology and Society), says that artificial intelligence has countless possibilities to contribute positively to education.

“Forbidding is not the way. It’s taking advantage of the technology’s potential, such as validating a text I’ve written, proofreading, helping with translation, helping with grammatical corrections, producing drawings during a class, trying immersive questions…”

Bill No. 2338, of 2023, addresses the creation of an indicative classification of the risks of artificial intelligence tools. It is a project, that is, it is not in force; but Junquilho draws attention to the high-risk classification of AI used in education.

“Considering that children and adolescents are in the educational and developmental phase, the AI ​​that is applied for education or for children and adolescents can be considered in Brazil by this bill as high risk, therefore it deserves more care”, he explains.

To be beneficial, such use depends on limitations and supervision by those in charge. “Generative artificial intelligences, those that generate text and image, can deceive very easily and bring a certain feeling of laziness in educational development, so why would you read a text if you can ask the AI ​​to summarize it for you”, exemplifies the researcher.

This is one of Ruschel’s concerns about his children: “The big challenge in using these tools is convincing a child that it’s just a tool and can’t do all the work for them.”

She is aware that stimulating learning beyond tools is necessary even so that children can make the best use of them, and she thinks about how she will transmit this learning so that her children do not become hostages of technologies.

“Even to realize that there is some information that is false or that does not make sense, to formulate the right questions in a search engine or in ChatGPT, all of this demands a minimum content and a reasoning capacity that does not come for free. It demands study and effort. of each, is something AI can’t do for you.”