Surprising as it may seem, parts of the UK national press have been buzzing with excitement over the terms of a non-disclosure agreement (NDA).
The story so far
There haven’t been any (many?) school exams this year, due to the Covid-19 pandemic. Instead, school pupils were to be awarded grades in their GCSE and A level exams using an algorithm developed by OFQUAL, the regulatory body for school examinations in England.
The Royal Statistical Society (RSS) had proposed to OFQUAL the inclusion of two independent experts to help develop the algorithm. OFQUAL asked these experts to sign an NDA. RSS objected to the terms of the NDA on the grounds that it would prevent RSS from commenting on the choice of computer model, on how the constraint of grade inflation was set, whether it was appropriate to the statistical task at hand, and who set it.
In fact, following an outcry over the results produced by the algorithm, pupils have been able to claim any higher results predicted by their teachers.
The terms of the NDA
So, what is so terrible about the NDA that was proposed? IP Draughts has reviewed many hundreds of NDAs over the last 36 years, and one of his books is on the subject of Drafting Confidentiality Agreements, published by the Law Society of England and Wales, and now in its third edition. He thought it might be interesting to review OFQUAL’s NDA.
First, some general comments, before focussing on the special case of an independent academic review of technology being used for public purposes.
IP Draughts’ overall impression is that the NDA was originally based on a simple template with conventional terms, but like many templates it has acquired some accretions over time, and it doesn’t look like it has received a high-quality legal review for some time. He would be embarrassed to put his name to such a scruffy document.
Points that jumped out (in the order in which they appear in the draft) include:
- Bad formatting, e.g. the definition of the Parties appears in a paragraph about OFQUAL.
- The definition of Confidential Information is broad. It covers all information provided by OFQUAL unless the information fits within one of the standard exceptions. More often in NDAs there would be a requirement that information must be marked as confidential or if disclosed orally identified as confidential at the time of disclosure.
- Bad/sloppy choice of language, e.g. “undertakes to do as follows” is followed by a list of items, such as “that copyright …remains the copyright of [OFQUAL]”
In a routine business-to-business (B2B) NDA, these points could be fixed by marking up the draft with conventional changes found in many NDAs, and with drafting corrections.
Other aspects of the NDA seem unremarkable in B2B NDAs, e.g.
- conventional exceptions to confidentiality
- restrictions on use or disclosure of the Confidential Information
- requirement to return information on request and certify destruction
- 5-year term for confidentiality obligations
Conventional as the above terms are, there are still a few drafting suggestions that might be made, e.g. that the receiving party can keep one copy of the confidential information in its legal files.
Special circumstances of this disclosure
OFQUAL’s NDA was proposed for signature by two independent experts suggested by the Royal Statistical Society. It appears from the RSS’s letters (see first link above and documents linked there) that they wanted to be free to publicly criticise the detailed features of the algorithm, and its creators, including OFQUAL. They were concerned that the NDA would prevent them from doing so.
IP Draughts agrees that the NDA probably would prevent them from doing so, but he has some sympathy with the OFQUAL position. It is a public body that was developing a policy, including a methodology for implementing the policy, on an issue affecting many thousands of families in England. As later events showed, the policy was very politically-sensitive.
In those circumstances, looking at it from OFQUAL’s perspective, it seems counter-intuitive to admit people to the policy-making body and then allow them to criticise the policy.
It appears that RSS is a campaigning body that presumably wants both better use of statistics in public bodies and greater transparency about their use. That is an understandable and laudable aim, but if I were running OFQUAL I would be reluctant to create a rod for my own back by giving RSS representatives both privileged access to a policy and freedom to criticise it.
Overall, IP Draughts thinks the issue is not really about the terms of a mediocre NDA, but about whether people involved in developing public policies should be constrained as to what they can say in public about those policies, and when. This is, or should be, itself a matter of public policy rather than something to be governed by private agreements.