Reflections from the final SE4RA project partner meeting hosted by TOBB ETÜ

The “publish or perish” culture has led, especially young academics, to utilise every possible tool to meet publication targets, including AI agents, which emerge as a disruptive tool for the whole research lifecycle. For AI-assisted outputs to be more resilient to the spread of misinformation, three safeguards- ethics review committees, adequate peer review, and the ethical robustness of the individual researcher are suggested to be updated. As a direct response, SE4RA took the self-evaluation instruments and tested them on real academic situations both in UCP and TOBB ETÜ. In order to provide a solid ground for these studies, a survey encompassing students, faculty, and academics was put underway that showed considerable AI use in these groups and a need for clear safeguards and proper assistance. From UCP, two methodological shifts emerged: first, the checklist being separated into two complementary parts to separate individual researcher responsibilities from the institutional ones; second, to allow the instrument to better function as an audit device, it has been changed into a more maturity-oriented model. SE4RA has gained significant momentum, although most of CoARA’s action plans are highly engaged but low on implementation. For the next steps, the consortium agreed to apply for COST Action funding to support networking activities, disseminate the self-evaluation checklist, reach additional institutions, and do a dissemination webinar on September 28.

Generative AI was identified as a disruptive factor for its abilities to perform tasks such as literature review, hypothesis framing, translation, and journal selection. This is presented as the direct motivation for accelerated regulatory and institutional attention: rebuilding trust in science and scientific evidence, and countering the spread of pseudoscience and misinformation.

Three safeguards for research integrity were identified, each only partially resilient to full delegation to AI: 

SE4RA was framed as a direct response to this third safeguard, an attempt to make the ethically sound choice the easy choice, at both the individual and institutional level.

SE4RA took the CoARA-ERIP ethics self-assessment instruments developed on paper by the CoARA Working Group and tested whether they could be operationalised inside two real institutions with different governance structures. The consortium comprised TOBB ETÜ and UCP as institutional testbeds, coordinated by the Research Data Alliance.

The top 5 most wanted guidance types were:

5% of academics and 34% of students stated they continued using AI in their work despite perceiving risks. Academics, when asked about in which circumstances AI use should be disclosed, answered;

Answers from the open-ended questions show clear hesitation and uncertainty regarding the boundaries of AI use in research. ALL of these findings point us towards clear guidance.

Two methodological shifts emerged from the UCP pilot as the project’s most transferable contributions. First, the checklist evolved from a single instrument into two complementary views: an institutional checklist (covering governance, integrity, fairness, transparency, AI-supported assessment, capacity building, and monitoring) and an individual-researcher checklist (covering open science, research data management, persistent identifiers, reproducibility, responsible AI use, and a qualitative-quantitative evidence mix). This separation was necessary because conflating the two risked transferring institutional responsibilities inappropriately onto individual researchers.

Second, the scoring model moved from a binary yes/no format to a maturity-oriented model, in which each item is considered together with supporting evidence and context. This way, the instrument can function better as a learning instrument rather than an audit device. It allows reflection and judgement instead of just generating a score that replaces institutional decision-making. 

UCP’s underlying safeguard message was that missing data in a research information system cannot simply be read as missing activity.  Before data is used for assessment purposes, institutions need validated definitions, clear provenance, proportionate data capture, interoperability, and explicit rules for handling missingness.

Despite this momentum, the consortium was explicit that engagement is not the same as implementation. The European Commission’s independent study on the Agreement of Reforming Research Assessment’s implementation (June 2026) found engagement with the Agreement broad and growing, but implementation inconsistent. An analysis of 248 CoARA action plans across 35 countries found most plans still in early developmental stages, with values-heavy language but limited concrete commitments. A related concern raised by Francis Crawley is that evidence-based assessment reform risks reproducing ranking systems in a new form rather than dismantling the underlying logic of ranking itself.

The Youth Health Council, although not an official partner, was present during the meeting, and through their collaborations with the Turkish Higher Education Council and related government agencies, they emerge as an important future partner for dissemination of the self-checklist and pilot implementations of research assessment into the university curriculum.

SE4RA’s central contribution is not just technical but methodological and cultural. It demonstrates that a shared self-evaluation instrument, developed once by the CoARA-ERIP working group, can be adapted to markedly different institutional governance structures while preserving comparability. Its two consortium partners showed that a bioethics-institute-led model and a newly constituted committee-led model can each produce meaningful institutional change when the emphasis is placed on evidence-informed judgement.