The idea is that UEs should emit the least power as possible. To allocate the biggest number of Subs in one cell.
In more details
The smaller the power, the less is the noise from others' point of view - my signal is your noise and vice versa. Performing the de/scrambling procedure, where noise level is low will give 1 or very close to 1 result in correlation procedure of scrambling codes. The higher the noise (from other Subs' signals) the higher probability/possibility that signal wont be de-scrambled to allow to decode the data. Because your signal will be gone in the noise.
Think about this like talking to someone next to you when being close to woofer on a ((death) metal) concert.
Additionally want to share the below.
Physics of electromagnetism with taste of modulation should answer your question. Long story short, weak signals are aligning to the phase of strong ones.