notes

Personal notes
git clone git://git.laack.co/notes.git
Log | Files | Refs

UnderstandingMentalModelsInAIDrivenCodeCompletion.md (1142B)


      1 # Understanding Mental Models in AI-Driven Code Completion
      2 
      3 **Source:** https://doi.org/10.1016/j.ijhcs.2025.103648
      4 
      5 ## Summary
      6 
      7 The researchers performed an elicitation study with 56 developers to evaluate developer preferences wrt how suggestons should be displayed, when they should be displayed, and other implementation details. 
      8 
      9 ### Thoughts as I Read
     10 
     11 - What if you had a complementary explanation of suggestions?
     12     - Perhaps this would lead to a hands off approach?
     13     - Maybe that wouldn't be part of the CCT, something additional
     14         - I could like that. A model is called with the AST and you ask a question like, "why is X assigned to Y here".
     15 - Two modalities
     16     - Acceleration
     17         - Speed up time to location
     18     - Exploration
     19         - Unsure where you are going
     20 - Should be possible to disable suggestions easily (toggle perhaps)
     21 
     22 ## Related
     23 
     24 - Research on productivity with AI-based code completion tools
     25     - http://dx.doi.org/10.2139/ssrn.4945566
     26     - http://dx.doi.org/10.48550/ARXIV.2302.06590
     27     - http://dx.doi.org/10.1145/3661145
     28 - [DORA Metrics](DORAMetrics.md)
     29 - [SPACE Framework](SPACEFramework.md)