There's a ton of research into this idea, have a search on PubMed. Unfortunately it suffers from at least two problems. First, generalisable effects from spec

There's a ton of research into this idea, have a search on PubMed. Unfortunately it suffers from at least two problems. First, generalisable effects from specific interventions often fail to replicate, so for example subjects tend to get better at a specific task but it doesn't necessarily carry over to real-world contexts. There are a few promising potential directions, although even those results are modest and uncertain. Second, people with even mild/early signs of dementia are extremely unlikely to continue with an intervention. Even if they do they often modify it, e.g. unplugging a smart device to save energy, or taping over buttons to prevent accidentally pressing them. I would hunt out references but I'm on mobile, they shouldn't be hard to find.