Is There a Way To Create a Chess Bot That Makes Logical Blunders and Plays Like a Human?

How can people create a Chess bot that doesn't play way under or above the desired rating and just plays a little above or below the standards for that specific rating bracket? I understand that it's difficult, but can or have anyone created a logical bot?

I assume it'd be easy with machine learning. You'd train the bot to guess the move. After it gets good at that you'd have a human-like player for that rating.

The maia bots are designed to do just that.  

from Maia Chess : 

Maia is an AlphaZero/Leela-like deep learning framework that learns from online human games instead of self-play. Maia is trained on millions of games, and tries to predict the human move played in each position seen.

We trained 9 versions of Maia, one for each Elo milestone between 1100 and 1900. Maia 1100 was only trained on games between 1100-rated players, and so on. Each version learned from 12 million human games, and learns how chess is typically played at its specific level.

liChess have three maia bots created from games between  1100, 1500, and 1900 players (liChess players).   They play rated games and their current rapid ratings are 1552, 1690 and 1776.

My proposal:

Set a certain "target centipawn loss".
Make the engine compute many lines, and choose the line that loses the closest to the target centipawn loss.

Example: Set target centipawn loss to 25.
Now, every move, the engine will try to lose as close as possible to 0.25.
This will make for an engine that makes many inaccuracies, but no big mistakes.

As for Maia, I found that bot worthless. She misses simple tactics and mates in 1. I hang pieces and still beat Maia 1900. I think they trained that thing on bullet games.