Let’s dive into the heart of a fascinating universe where artificial intelligence takes over surveys to shape our perception of the world. What will happen when these machines take the reins of our opinion polls? Come discover the implications of this technological revolution on our choices, our decisions, and our society.
The question may seem futuristic, even far-fetched, but it is indeed current. Could artificial intelligence (AI) replace humans in polls and sociological surveys? As language models like ChatGPT and Copilot continue to improve, researchers are exploring new avenues to leverage their potential. Before jumping to conclusions, let’s take a closer look at what this entails.
Technological advances in question
Language models (LLM) like ChatGPT, developed by OpenAI and other companies, can understand and produce text impressively. These algorithms can imitate specific socio-economic profiles and answer questions while posing as different categories of the population. A study conducted by Brigham Young University in 2023 demonstrated that the responses provided by these AIs for fictional panels could closely match the voting distribution in the United States across several elections.
Potential advantages
Using AI to conduct surveys has several undeniable advantages. First, data can be collected more quickly than with traditional methods. Secondly, “silicon samples,” as some researchers call them, can reduce common biases in human studies, such as low response rates or discrepancies in participant recruitment.
Promising but imperfect results
Experiments have shown that, under certain conditions, AIs can indeed reproduce human behaviors in surprisingly faithful ways. For example, in the fields of economics and cognitive psychology, models have successfully imitated the choices and opinions of humans in classic experiments.
The limits of artificial intelligence
Nevertheless, these advances are not without their limitations. A recent study conducted by a French team from the Center for Research in Economics and Statistics at École Polytechnique revealed that AIs do not always shine when comparing their responses to a global benchmark database. The models tested by OpenAI, Meta, or Mistral showed inconsistencies when it came to delivering responses on political opinion, happiness, religious practice, and trust in others.
Ethical and societal questions
In addition to technical challenges, the use of AIs to conduct surveys also raises ethical and societal questions. What happens if AIs demonize certain responses? How can anonymity and confidentiality of “participants” be ensured when it is machines that respond? And finally, what is the impact on the employment of professionals in social science research?
The future of surveys and sociological studies
Despite the challenges, it is clear that artificial intelligence has enormous potential to reshape how we conduct surveys and sociological studies. Researchers will need to continue refining these technologies while addressing the ethical questions their use raises. One thing is certain: the era of “silicon samples” has only just begun.
