Walk the Walk: Promoting Research Integrity

29 January 2026 27 January 2026
29 January 2026

Context

The overarching aim of the project is to improve the practice of research integrity across the planet. The current situation for three categories of integrity is reviewed. A small survey of manuscripts in Scientific Reports illustrates the gap between practice and declared aims. Plans to produce, test and disseminate tools are described, with current resources. The conclusion asks for collaborators and testers.

Current State of Research Integrity Practice

There are three main components of research integrity: openness, risk, methodology. All are supported by agreed international guidelines.

Open: Publications & Data

  • Author’s responsibility. Owns intellectual property rights (IPR),
  • Public email for personal requests & public repository DOI. Beware GDPR,
  • Journal responsibility. Available: immediately GOLD or after interval GREEN,
  • Data statement, author identity, risk (ethics).  Follow declared commitments,
  • Charge a bomb, Article Processing Charges (APC) or subscription,
  • In my humble opinion, should be accountable & pay substantially for breaches of commitments,
  • Probably often breached. Personal experience, exploratory study,
  • Sponsor responsibilities. Distribution of income according to declared commitments,
  • Details of awarded proposal,
  • Probably often breached. Personal experience.

Risk: Ethics, animal rights general,

  • External approval: organisation, protocol number.

Rigorous methods: equipment details, statistical methods

  • Measurement unit, variables, with selection criteria, descriptive, inferential statistics.

Exploratory Study

Further details available

Method

I enlisted the help of ChatGPT and trotted off to Scientific Reports and arbitrarily downloaded numbers 1, 4, 8, and 11 of the most cited psychology manuscripts of 2024. I chose 4 and 8 because they administered stuff to people and I was interested in risk to participants, number 11 as it was purely behavioural and number 1 because it was number 1 (Barba et al., 2024; Cuttler et al., 2024; Gordji-Nejad et al., 2024; Stein et al., 2024).

I checked manually against Nature’s Editorial Values: data for ‘openness’ and ‘transparency’;  ethics for ‘highest standards in research culture’; methodology & results for ‘rigour’ and ‘reproducibility’. Also used ChatGPT. NB - work already published, so no IPR issues.

Data deposit results

Manually, 3/4 manuscripts. There was no recognised depository. Some stated ‘data was available from corresponding author on request, with or without public email. Hope none of these authors fall under a bus. Think the grieving heirs could locate the data?

In response to “check all 4 manuscripts for data deposit compliance with FAIR and OpenAIRE” gave: FAIR 1@4/4, 2@2/4, 1@1,4 (survey); OpenAIRE 1 yes, 2 partial, 1 no (survey).

Ethics Results

Better. The 3 in psychology departments named ethics committee, not the survey. All had approval and consent. None provided patient information sheet, or risk, or privacy.

Methodology results

Manually, no manuscript explicitly reported normality checks or outlier handling. The p-value was exact for 2 manuscripts, variable for 2. No study reported effect size.

ChatGPT answered to “Methodological rigour, are method and results sufficiently detailed for replication”. Method: 1 High procedural clarity, 1 very detailed, 1 complete, 1 moderate. Data access: 2 partial (no raw), 1 restricted (on request), 1 full OSF. Code 1 full, 1 standardised 1 partial 1 none. Overall: 1 excellent, 1 strong 1, good, 1 adequate

ChatGPT rated question “excellent”. Not sure I can return the compliment.

Results summary. Scientific Reports practices research integrity? You cannot be serious.

Plans

Promulgate existing resources.

Create tools. Initially for Journal authors, editors and reviewers. Then for sponsors.

Expand the text check list already available.

Create a spreadsheet version of check list.

Create AI version that can be used without uploading to internet.

Test tools. Minimum 2 highly cited 2024 manuscripts from  journals. Selection is resource dependent. Hopefully will include:

  • General Nature Human Behaviour, Scientific Reports,
  • Cognitive, Social, Clinical. UK, EU, USA. Most cited 2025.

Author

Emeritus Professor in Mathematical Psychology, Diana Kornbrot

Diana welcomes collaboration in this area from academic colleagues both internal and external to the University of Hertfordshire, journals, charities, learned societies, and/or any other relevant research organisations.