One session used hypothetical evidence-room discrepancy.
Prosecutors asked:
Can we rehabilitate?
Defense asked:
Can we exclude?
I asked:
What happened?
They hated me.
Eventually they saw point.
Before deciding legal strategy, understand facts.
A defense lawyer told me:
“You say that because you don’t have client.”
“Correct.”
“I have job to attack chain.”
“Yes.”
“And prosecutor defend.”
“Yes.”
“So neutral fact-finding is fantasy.”
“No.”
“It is different role.”
That distinction matters.
Advocacy is legitimate.
Audit is different.
A good justice system needs both.
If everyone is advocate, facts become weapons only.
If everyone claims neutrality, power hides.
Roles.
Boundaries.
Again.
At one training, a prosecutor who had worked Harland case attended.
Older now.
Afterward he said:
“You remember when I wanted you to say jacket was planted?”
“Yes.”
“You were right.”
“I know.”
He laughed.
“Still impossible.”
“What changed?”
“Trial.”
“How?”
“Defense would’ve destroyed word planted because we couldn’t prove physical placement.”
“Yes.”
“We didn’t need it.”
“No.”
He looked.
“I think about that now.”
“When younger attorneys overcharge theory.”
“Good.”
He smiled.
“You’re taking credit?”
“No.”
“I am.”
We had coffee.
He told me former sergeant had requested early release once and been denied.
I asked no details.
That part no longer mine.
The prosecutor said something interesting:
“I hated him during trial.”
“Did you?”
“Yes.”
“That worried me.”
“Why?”
“Because I was making decisions.”
Good.
“What did you do?”
“I assigned another prosecutor to review charging choices.”
“Excellent.”
He nodded.
“I learned from you.”
“That sounds dangerous.”
He laughed.
Even prosecutors need second eyes when emotion grows.
So do auditors.
So do judges.
No one gets exemption.
By seventy-two, my knees objected to airports.
I switched mostly remote review.
Technology improved.
Digitized evidence made some audits easier.
Hard drives retained histories paper never could.
But new vulnerabilities appeared.
Access rights.
Metadata manipulation.
Cloud synchronization.
Automated retention.
Algorithmic classification.
One department used AI system to flag chain anomalies.
Officials treated score as authority.
I reviewed.
The model disproportionately flagged cases with manually entered historical records because missing structured fields looked suspicious.
Those cases often came from older precincts serving lower-income neighborhoods.
Suddenly an “objective” system generated apparent corruption clusters along socioeconomic lines.
No malicious coder required.
Data structure.
We corrected.
That case made me uneasy.
People wanted technology to replace judgment.
It usually relocates judgment.
Someone chooses training data.
Threshold.
Definition of anomaly.
False-positive tolerance.
Who gets reviewed.