Izjava o upotrebi veštačke inteligencije
Gde Fidestat koristi veštačku inteligenciju, koje podatke koristi i šta ne radi.
- Verzija:
- 1
- Stupa na snagu:
Ovaj dokument je dostupan samo na engleskom jeziku. Važeći je tekst na engleskom ispod.
This statement explains where Fidestat uses artificial intelligence, what data it uses and what it does not do.
1. Where we use AI
Survey generation: you describe your goal in a chat and an AI model drafts the survey — questions, answer options and order. You review and edit the draft before publishing.
Reports: an AI model writes the text of the research report — findings, interpretation and limitations — from results calculated by our software.
Advisor and recommendations: an AI model suggests business actions based on the findings, and answers your follow-up questions about them.
2. Where we do not use AI
The interviewer that respondents talk to follows the survey exactly as you designed it. It does not use a language model to change questions, wording or order. Respondents are told before the first question that they are answering an automated interviewer.
Statistics, percentages, confidence intervals, charts and tables are calculated by our software, not generated by AI. The AI model receives these numbers and writes about them; it does not invent them.
3. Our methodology
AI output in Fidestat is guided by a proprietary methodology built from established research literature on survey design, statistics and interpretation. It shapes how questions are drafted and how results are interpreted. It does not replace your judgement.
4. Data used
Survey generation: the text you write in the AI chat and the survey task you define.
Reports and Advisor: the survey title, aggregated results, and short quotes (up to 300 characters) of free-text answers. Respondents' text is passed to the model only as data to analyse, never as instructions.
AI features run only when you use them. Free-plan accounts do not send any data to an AI provider.
5. AI provider and training
We use models from Anthropic (see the Sub-processor List). Under Anthropic's commercial terms, the data we send is not used to train its models. Fidestat does not use your surveys or responses to train any AI model.
6. Limits of AI output
AI output can be incomplete or wrong, especially with small samples. Reports state their limitations, and findings based on small groups are marked as indicative. You remain responsible for reviewing AI output and for decisions based on it. AI output is never used to make automated decisions about individual respondents.
7. Contact
Questions about our use of AI: Contact Fidestat at https://fidestat.com/contact