Fridae

Audrey Tang takes an inclusion question into the software-buying process in a new Open for Business report on Taiwan. Looking back at marriage equality in 2019, the cyber ambassador describes how databases adapted when their fields hadn't already ruled out different kinds of families.

Tang's advice is to deal with that design early: “schemata are cheapest to fix before procurement, not after deployment.” In other words, look at the structure of the records before buying the system.

For recruitment software, a name change is one place to start. Make up an application with a current name and a qualification under an earlier one. Ask the supplier to carry it through the process. Can the name be corrected? Which version does the interviewer see? There's no need to use a colleague's history for the demonstration.

The same conversation can cover a gender field or a question about a partner. Why is it needed, who reads it, and can it be left blank? For LGBTQ+ applicants, the answer may involve information they haven't chosen to share at work. A form that handles a name change still tells you little about the fairness of a scoring model. Ask which data goes into it, what backs the supplier's claims, and who can deal with errors.

The UK's Information Commissioner's Office raised supplier scrutiny, minimal personal data and explanations to candidates in November 2024 guidance on recruitment AI. That is UK guidance, not Taiwanese law. Buyers elsewhere can use the questions while keeping the jurisdiction clear.

Keep the supplier's answers with the purchase. For applicants, an explanation of automated assessment and a contact for questions are useful things to look for before giving extra details.