EC-16.5 Standards, and Research Integrity

What this is and why it exists

Standards and research integrity sit in one topic because both answer the same question.

What makes an engineering statement checkable by someone who was not there?

A standard makes a product statement checkable, by fixing what the words mean and what test demonstrates them. Integrity makes a research finding checkable, by fixing what was actually done and who did it. Treating them together is unusual, and the connection is the argument this topic makes.

The vocabulary

• Standard — an agreed specification produced by a recognised body.
• Normative reference — another document a standard obliges you to satisfy.
• Conformance — meeting a standard's requirements.
• Certification — a third party attesting that conformance was demonstrated.
• Attribution — crediting the source of work you have used.
• Fabrication — inventing or altering data.
• Authorship — the set of people credited with producing a piece of work.

The mental model

A standard is an agreed specification produced by a body whose members are industry, academia and sometimes government.

That membership explains both its authority and its compromises. A standard is a consensus document. It carries weight because many parties agreed to it, and it contains provisions that exist because somebody would not agree without them. Neither fact should surprise you when reading one.

Standards are long, heavily cross-referenced, and written to be searched rather than read through. The skill is locating the clause that applies and then following its references.

The references matter more than they look. A normative reference is a document the standard obliges you to satisfy, so conforming to one standard can require conforming to three. An informative reference is background you may ignore. Confusing the two is how a project discovers a missing requirement late.

Three words are used loosely and mean different things.

Conformance is meeting the specification. Testing is demonstrating that you meet it. Certification is a third party attesting that the demonstration happened. A product can conform without being certified. The distinction is commercial as much as technical, because certification is often what a customer or a market requires.

Now integrity, and the connection. Attribution is crediting the source of code, figures, text or ideas that you used, whatever the licence permits.

The licence and the obligation are different questions. A permissive licence allows you to use code; it does not make it your code. Presenting someone else's work as your own is the one failure that ends careers rather than damaging them. It is judged by what was represented, not by what was legally permitted.

Handling data honestly has one rule with one exception. Every point is reported, or the exclusion is stated with its reason.

That covers all three of the common failures. Removing inconvenient results is fabrication. Adjusting a figure to look cleaner is fabrication. Reporting the best run as though it were typical is fabrication, even when that run really happened. What is represented is not what was observed.

Authorship reflects contribution, and it should be discussed at the start of a project rather than at the end.

Someone whose work is used is credited whether or not they are present to object, and whether or not they are still on the project. Deciding this early costs a conversation and deciding it late costs an argument. The argument is usually between people who each remember their own contribution most vividly.

What you should now be able to explain or do

• Say what a standard is, who writes one, and why it contains compromises.
• Search a standard for an applicable clause and follow its normative references.
• Distinguish conformance, testing and certification, and say why the difference is commercial.
• Attribute work you have used, independently of what its licence permits.
• Apply the reporting rule and recognise the three common forms of fabrication.
• Agree authorship at the start of a project rather than at the end.

Check yourself