The ITA team uses artificial intelligence (AI) responsibly and transparently. We regard AI as a tool that supports us in all aspects of technology assessment, but which cannot replace us. In accordance with good scientific practice, we continue to take full responsibility for the quality of our findings and reports. The recipients of our research in society and politics can rest assured that we not only comply with the law but also approach the use of AI in a considered manner in many respects, and that people remain the authors of the ITA technology assessment reports.
The following eight AI-related guidelines guide our work:
In detail, these eight points mean:
1. Compliance with laws and regulations
When using AI tools, we adhere to all legal requirements, the rules laid down by the Austrian Academy of Sciences (ÖAW) and by specific clients or publishers, as well as the rules of good scientific practice, both in general and within specific (disciplinary/methodological) fields of application.
We use only AI tools that comply with the law and, as far as can be ascertained, are technically secure. We give preference to AI tools that have been developed by, within, and for the academic community, are available via open-source models, and/or can be hosted locally.
2. Quality control
The use of AI tools at the ITA will be carried out in a considered and responsible manner and, until further notice, only with active, ongoing quality control: this means that we constantly review the output of the AI tools and ensure that we possess the necessary knowledge for careful data entry and the evaluation of the results. We remain responsible for the scientific quality and societal robustness of our work.
3. A Critical Approach
We are aware that AI is not currently a reliable technology. Results from AI tools can be flawed and misleading (e.g. hallucinations, algorithmic bias). Furthermore, the underlying sources may contain factual inaccuracies and, in some circumstances, lead to unjustified bias. Furthermore, we anticipate that our research will increasingly uncover documents that have themselves been generated by AI tools. Where verifiable, the source reference (existence, quality, and representativeness of the sources) will therefore always be considered. In accordance with the principles of good scientific practice, AI-generated results are, as a rule, only adopted after verification.
4. People first
We do not use AI as a substitute for dialogue, internal consultation, and creative thinking. We remain our own best sparring partners. We draw on the multi- and interdisciplinary expertise of the Institute for creativity, data collection and quality assurance. We do not delegate tasks entirely to AI tools; human intelligence and direct social interaction continue to play a decisive role in all areas.
5. Prudent use of AI
Due to the negative environmental and social impacts of current AI infrastructure, we use AI tools only in a targeted, supportive, and sparing manner. AI tools do not serve as a substitute for traditional methods that are equally efficient, more resource-efficient, and more socially acceptable. We use AI tools only if they genuinely support the process to be carried out and make a significant positive contribution to the quality of the results.
6. Transparency
We state in advance, both internally and externally, when and for what purpose we are using or have used AI tools. In doing so, we distinguish between frequently used tools and specific applications for content creation. The former are listed as examples on the ITA website, whilst the latter are explicitly mentioned in the relevant section of the respective report. Furthermore, we also require this transparency from our collaboration partners and co-authors.
7. Learning process and continuous evaluation
We continually reflect on the latest developments in AI tools and their potential applications, participate in the relevant debates within our international community, and adapt our approach to AI where necessary. As AI is also the subject of our research, the insights gained from various application contexts are incorporated into our ongoing reflection and practical adjustments within TA.
8. Specific features of social and policy consultancy
To ensure that TA results are socially robust, it is necessary to continually and critically reflect on the wording used when employing AI, so as not to adopt AI-generated language, framing, and bias unthinkingly. Similarly, we exercise caution when selecting topics and issues suggested by AI-supported processes, as these too may be subject to bias.
