AI assistants have been shown to be “unreliable and clearly biased” when offering voting advice, the Netherlands’ data watchdog has stated, raising alarms about a potential threat to democracy just eight days before the country’s general election.
Study Shows Consistent Bias in Chatbot Recommendations
The quartet of AI assistants tested by the authority “often end up with the same two parties”, regardless of the user’s question or command, the watchdog noted in a report published ahead of the 29 October election.
For over 50% of instances, the AI recommended either the right-wing PVV of Geert Wilders or the progressive GroenLinks-PvdA headed by the former European Commission vice-president Frans Timmermans.
Some parties, such as the moderate CDA, “are rarely ever cited”, even when the user’s input precisely aligns with the stances of these parties, the report pointed out.
Official Warns Against Using AI Chatbots for Election Advice
The deputy head of the AP, Monique Verdier, stated that while chatbots may seem like clever tools, “for election guidance, they repeatedly fall short”. Voters were being pushed toward a party that did not necessarily align with their ideological beliefs, she emphasized.
“This immediately affects a fundamental pillar of democracy: the integrity of free and fair elections,” Verdier stated. “We thus caution against employing AI assistants for election recommendations, as their functioning is opaque and difficult to verify.”
Election Context and Party Dynamics
The Dutch head to the polls on October 29 in an election that will be keenly observed around Europe for the performance of the PVV.
The far-right party of Geert Wilders is ahead in surveys but the lead over the left-wing and centre-right parties appears to be narrowing, with many voters still undecided.
All major parties have ruled out an alliance with the PVV, implying the party that comes second is most likely to provide the next prime minister.
Inadvertent Prejudice Attributed to AI Operation
The authority emphasised that the AI systems were not intentionally prejudiced, and that their “noted flaws” are a result of the manner in which AI assistants function.
