ChatGPT - Scenario writer

Traditional festivals in India frequently involve firecrackers and parties, increasing levels of air and water pollution and health risks. Though street theatre is acknowledged as a great instrument to motivate behavioural change, neither FIN’s staff nor its mentors felt capable of writing a street performance. However, within three minutes of being fed the right prompt, ChatGPT came up with a delightful skit involving youths. It included both male and female characters, used local names and was mindful of local nuances. It argued that the impact on our oceans of microfibres from synthetic clothing represents a significant environmental problem that would ultimately affect livelihoods adversely. The FIN staff were delighted and boosted the local character of the skit by inserting jokes.

ChatGPT a hallucinator

Women’s empowerment is a central mission of FIN. In a second experiment, we asked ChatGPT to create a speech explaining the relevance of International Women’s Day to villagers where FIN operates. The speech was very impressive, but it contained incorrect information on sex ratios, the abortion of foetuses outside legal limits, and women’s participation in the workforce. When ChatGPT was asked to justify the errors, it replied: “I apologise for any confusion. I provided a hypothetical statistic to illustrate the point in the pitch. It is always recommended to use accurate and up-to-date statistics when making a pitch or argument.”

ChatGPT a ready advisor – but too confusing

In the third experiment, we taught the FIN ladies staff themselves to use ChatGPT. They were able to install it on their phones easily. They found it incredibly helpful that unlike with Google, they could hold a conversation with ChatGPT, making it like a real friend.

One of them wanted to have a debate with her husband about a candidate in the local state elections. So, she asked ChatGPT what was good about her preferred candidate and what was bad about the opponent. The answer was satisfactory, but when she did this for the opponent, the response was equally convincing. As FIN staff are not skilled enough to check the veracity of arguments through cross-checking research, the responses from ChatGPT made her more confused. So, her verdict on ChatGPT was: “No thank you, not for me”.

ChatGPT too useful for rabble rousers

In a fourth experiment, we went into the dark side. Even today, India remains plagued by communal violence, honour killings, and other forms of violence against those who challenge entrenched social norms and the status quo. We conducted additional experiments and discovered that ChatGPT could generate compelling, emotionally charged arguments in seconds to justify such vicious actions and also to challenge them.

To ensure the accuracy of our results, we asked ChatGPT for its opinion.  It asserted that ChatGPT can be a valuable tool for both economically disadvantaged people and NGOs by providing information, offering emotional support and making communication more effective. But the chatbot avoided discussing its obvious downside, namely hallucination i.e. effectively arguing points based on false, incomplete or imperfect information.

ChatGPT for BOP or NGOs?

So, for the time being, ChatGPT seems a handy tool for well-intentioned NGOs, but not useful for the ordinary BoP individual. Further, it can be a free and excellent script writer for violence-mongers aiming to raise social tensions among the BoP. Indeed, without the means to monitor the ethics and truthfulness of its friendly conversations, ChatGPT and similar AI systems could become the loudest voices in the already deafening echo chambers of opinion. It could become a dangerous tool if used by those unwilling or unable to rigorously fact-check. Is the risk of creating a million more Trumps at the BoP worth it?