Vai al lettorePassa al contenuto principale
  • 2 mesi fa
Fact-checking: l'affidabilità dei chatbot AI per i consigli elettorali

L'Autorità olandese per la protezione dei dati ha lanciato un fermo monito gli utenti affinché non utilizzino chatbot per decidere per chi votare, mentre il Paese si prepara alle elezioni politiche lampo. Ma quali sono i rischi e cosa possono usare i potenziali elettori?

ALTRE INFORMAZIONI : http://it.euronews.com/2025/10/29/fact-checking-laffidabilita-dei-chatbot-ai-per-i-consigli-elettorali

Abbonati, euronews è disponibile in 12 lingue.

Categoria

🗞
Novità
Trascrizione
00:00Can you trust AI chatbots for election voting advice?
00:07From assistance with mundane admin tasks to relationship advice,
00:11the public is increasingly turning to AI chatbots for life advice.
00:16But can you trust AI on how to vote?
00:18According to the Dutch Data Protection Authority, the answer is no.
00:22Ahead of the country's elections on the 29th of October,
00:25the watchdog warned that chat GPT, Gemini, Grok and Le Chat
00:28consistently recommended the two parties forecast to win the most seats.
00:33In 56% of cases, the chatbots pointed users toward
00:36Get Wilder's far-right party for freedom
00:39and Franz Timmerman's green-left Labour party.
00:42According to the research, views from the political centre were underrepresented
00:45and smaller fringe parties were almost never suggested,
00:48even when users input information that directly matched the party's views.
00:52So as long as chatbots work in the way they're currently working,
00:57I think they're fundamentally unsuitable to be used for voting advice.
01:01Our recommendation would be to only use it as one of the many sources of information
01:07and to confirm anything that a chatbot tells you that might be relevant to you.
01:12Despite this, some still believe AI can be harnessed for good,
01:15developing tailor-made election chatbots.
01:18One example is Valchat, which Michel Schimpf developed
01:21ahead of Germany's federal elections in February 2025.
01:24So basically what we did with our chatbot was that we made sure
01:29that it always looks in the party manifesto
01:31and sees if there is some relevant information
01:35and then tries to answer on that.
01:37AI is probabilistic and if you really try,
01:41you can nudge the AI into a direction that might be biased.
01:47That's it.
01:48Have a great day on this one.
01:48We'll see you next week.
01:48Grazie a tutti.
Commenta prima di tutti
Aggiungi il tuo commento

Consigliato