In March, the STRIX AI Awards competition, organized by Kreatív publisher PPH Media, concluded, recognizing the most innovative artificial intelligence solutions of the year. The joint submission by the Minerva Institute and Clementine, titled “The One Asking the Questions: ‘Robot Girl’”, addresses the challenges of telephone-based public opinion research. The project received a silver award in the “other” category.
The “The One Asking the Questions: ‘Robot Girl’” project offers an innovative response to the fundamental challenges of the CATI (Computer-Assisted Telephone Interviewing) industry. The sector faces issues such as interviewer bias, high cost and time requirements, declining response rates, as well as a shortage of human labour and inherent subjectivity.
Developed by the Minerva Institute and Clementine, the AI-driven telephone interviewer named Nóra is built on Clementine’s natural language processing platform, which has already proven itself in the market (the nationwide MÁV Elvira voice assistant is also based on this technology).
Nóra is Hungary’s only fully AI-based telephone survey interviewer system. The AI interviewer asks questions in a human-like voice and clearly indicates at the beginning of the conversation that it is an AI. It remains completely objective, reads out the questions exactly as written, and records all responses without bias. It is capable of processing answers in real time, and uses modern large language models to interpret and structure even open-ended, free-text responses. The solution simultaneously ensures objectivity, speed, scalability, and cost-efficiency.
The project is built on the Clementine team’s multi-year expertise in text analytics and speech-based AI solutions. A key component is the existing and proven Clementine Hanga platform, on which the survey-specialised AI assistant Nóra was developed. Since 2023, Hanga has supported the voice-based passenger communication system of MÁV Elvira.
The implementation phase of the project began with the launch of the Minerva Institute in spring 2025, followed by system testing and fine-tuning throughout the same year. Pilot studies confirmed both the system’s functionality and its acceptance. The project is currently under continuous development, with future phases focusing on further optimization of component-level interoperability and the expansion of system functionality.
Impact of the Process
Objectivity and reliability:
The most significant impact is the minimisation of both direct and indirect human influence. Nóra remains fully objective, does not suggest answers, and reads out the questions exactly as written to every respondent in a consistent manner, thereby increasing the reliability of research results. Respondents have also appreciated this robotic operation, with many highlighting as a positive aspect that they were able to express their opinions without judgement.
Speed and efficiency:
The process has been significantly accelerated. Collecting and processing a representative sample of 1000 respondents takes only 2–3 days, which is considerably faster than traditional methods. The system is capable of making calls and receiving callbacks 24/7, every day of the week. The greatest time savings are observed in data analysis: processing responses to open-ended questions, which previously took days, is completed within minutes, as the system immediately interprets and categorizes responses.
Scalability and flexibility:
Depending on the research objectives, the system can flexibly deploy 50, 100, or 500 “trained” robot interviewers operating at a standardised quality level. In addition, the AI system is capable of continuously dialing randomly generated numbers until a demographically representative sample is achieved. If required, it can also initiate callback attempts.
Transparency:
The efficiency of the process enables the Institute to publish anonymized raw response datasets immediately, ensuring transparency and allowing any interested party to conduct further analysis.
Measurable results
The Minerva Institute was established in spring 2025 and is currently primarily conducting its own pilot studies in order to build market awareness and demonstrate the potential of AI-based telephone public opinion research. Over the past months, the project has generated multiple media appearances, and several research institutes have reached out to the organisation regarding future collaboration and joint work.
The primary objective of this early phase is to familiarise the professional community with the Institute’s innovative solutions, thereby strengthening its reputation and laying the groundwork for long-term business and research partnerships. This reputation-building effort enables the Minerva Institute to rapidly and credibly position itself in the market, supporting future financial and commercial success.
A key priority of the Minerva Institute is to ensure that research results are transparent and accessible, and that their public availability is guaranteed for society as a whole. In doing so, the Institute aims to contribute to the development of a more democratic formation of public opinion and to strengthen societal dialogue.
(The Minerva Institute and Clementine were joined in the pilot studies by Wozawez Consulting, SpeechTex Ltd., and the Department of Telecommunications and Artificial Intelligence BME TMIT)