Riot has only ever briefly experimented on the PBE with negative scaling, but never in the form of items. What does the community think of players being able to opt into more specialized item paths based on the idea of actively giving something up to acquire something better? If someone wanted to double-down on durability at the cost of actively losing damage, or gain attack damage at the cost of movement speed, wouldn't that option provide for more varied builds from game to game and make each game more interesting?
In case the idea isn't clear, let me create a few hypothetical items to demonstrate (note: not saying these items are balanced).
Obsidian Crest: 100 Armor + 800 Health + Passive: Reduce all damage you deal by 25% (Limit to 1 per player)
Arcane Seal: 90 Magic Resist + 500 Health + 150% Health Regen + Passive: Increase all physical damage received by 15%
Resonance Sabre: 100 Attack Damage + 20% Critical Strike Chance + Passive: The duration of all crowd control effects on you are doubled
Items like this (in theory) would allow for more distinct build paths, so champions/items don't have to balanced around the possibility of all champs buying all items. If a juggernaut or assassin wanted to buy an incredibly tanky item, they'd lose the damage potential that they currently can have while building tanky.
A game in which a full-tank top lane gets multiple Infernal Dragons could allow them to invest in an item that reduces their damage output without having it dip noticeably below their base values. Likewise a tank fighting against a team that secures multiple Infernal Dragons could sacrifice their damage output since they're not expected to deal much anyway. If a carry has a team that can adequately protect them, they can dip into extra damage at the cost of extra vulnerability if someone manages to jump on top of them.
Not every negative stat needs to be something incredibly crippling, but as someone that craves diversity in both champions and items, I feel like this unexplored space has plenty of potential.