Good question! It made me do some research.
Compare decide and decision:
- decide → [dɪˈsaɪd]
- decision → [dɪˈsɪʒn̩]
'Decide' has a long vowel, 'decision' has short. 'Decisive' also has a long vowel/diphthong even though both 'decision' and 'decisive' are trisyllabic:
Here's my best shot:
According to this answer on ELU, the suffix -ion must have been disyllabic in the past:
- [di.ˈsi:.zi.ən] (Middle English)
So there are two syllables after the stressed syllable having a long vowel, meaning it's a prime candidate for Trisyllabic Laxing! Trisyllabic Laxing (TSL) was highly productive in Middle English, so it must have been applied to 'decision', shortening the vowel in its second syllable:
- [di.ˈsi:.zi.ən] → [di.ˈsi.zi.ən] (i: → i)
The Great Vowel Shift didn't affect it because it already had a short vowel.1 (Note, however, that the MidEng TSL applied before the GVS).
Now if you have noticed, there's a glide [j] when we move from a front vowel to another vowel:
Later on, the [z] and [j] coalesced2 to [ʒ], yielding [dɪˈsɪʒn̩]
Trisyllabic Laxing didn't affect it because it only had one syllable after the stressed syllable. I don't know much about the history of 'decide'. It was /dɛːˈsiːdən/ in Middle English. The unstressed schwa and terminal nasal were lost in Middle English.
It would've been [dɪˈsiːsiv] in Middle English. TSL didn't apply because it only had one syllable after the stressed syllable. Then after the Great Vowel Shift, it became [dɪˈsaɪsɪv].3
I explained it very briefly; there's an awful lot of history to it and I'm still sceptical about my answer. Colin Fine is probably who I'd ask about it.
1. GVS only affected long vowels
2. It's palatalisation
3. The GVS changed [iː] to [aɪ]