Testing and things going wrong
What needs checking before people can use it?
Can a blocked person still send a message? Where does private information go? What happens if the server stops? These pages explain the practical questions behind the testing plans.
We'd start with fictional users. The checks described here are work to be done before opening a real service.
Would a Block work straight away?
If I Block someone, could an old suggestion let them back in?
It mustn't. A Block has to apply when the person tries to see your profile or contact you, even if they were suggested earlier. The same goes for a photo you've stopped sharing or an account that's been restricted.
Why might some things take a little longer?
Preparing suggestions, checking media and making an optional summary can take time. They could run in the background while you use the rest of the service. If you edit your profile during that work, an old result mustn't replace your new words.
What if the AI is busy or unavailable?
An optional summary could fall back to your original text. A required safety check couldn't simply be skipped to make things faster. The application is still unbuilt; these are the rules we'd need to test before opening it.
Where would my information go?
Would all my information end up in one big pile?
It shouldn't. The plan keeps your legal identity separate from everyday dating information. Profiles, private conversations and safety reports would each have their own limits on who can see them and how long they're kept.
Would you sell my dating information or send it to outside AI?
Selling dating data isn't part of the plan. Neither are advertising trackers or routine outside analytics. AI work would normally run on the service's own systems. Any exception for outside processing would need a limited purpose and appropriate permission.
Peter's reasons for building it are on the About page.
Could I read a profile in another language?
Optional translation is planned, with the original still available. AI translations and summaries are aids to reading, not new facts about the person. If the original changes, the old version must stop being shown until it's updated; a profile summary would also need the owner's approval.
Would technical records contain my conversations?
Ordinary records of errors and performance shouldn't contain raw profiles or chats. A specific safety report is different and would have its own access and retention rules. Those details still need to be proved in the future application.
How would you test it?
How do you test a dating site without real members?
We'd start with fictional people and profiles created for testing. They could send messages, change their minds, Block each other and try to break things. That lets us check awkward situations without asking people to put their real dating lives into unfinished software.
Could real people help test it?
Yes, later. They'd preferably make up profiles and try things such as finding a setting or starting a conversation. We would need clear instructions and data rules before inviting anyone. Those tests would be separate from any future live dating service.
What would the test MitID button do?
It would let a fictional account pretend to complete registration so the rest of the journey could be tested. It wouldn't contact MitID or verify anybody's identity. The test version would have to be clearly labelled and kept out of the live service.
Does passing the tests mean it's ready to launch?
No. We'd still need to choose the first release, complete the remaining safety and operating checks, and connect real MitID after the other functions work. There's no public application or open testing programme here yet.
Before opening, I'd also need to establish the company that would operate the service and settle its privacy, support and legal responsibilities. Forming a company doesn't by itself make the service compliant.
What if something stops working?
What would happen if the service went down?
You should get a clear explanation of what's unavailable. If an optional summary stops working, you might still read the original profile. If the service can't check whether you're allowed to see something, it must stop that action rather than guess.
Could an outage undo a Block?
It mustn't. Blocking and taking away access need to work immediately whenever an action is accepted. A slow suggestions list can't keep a door open that you've just closed.
Could you restore lost information?
That would need to be demonstrated before launch. Having a backup file is only the beginning: we'd need to restore it on a clean system and check the accounts, conversations, photos and privacy choices together. We aren't making promises about recovery times before those tests have been done.