Page 1 of 1

Fingerprints and the IAI (Inquisition of Artificial Intelligence) - A ChatGPT thread

Posted: Fri Mar 17, 2023 10:07 am
by Boyd Baumgartner
Feeling overwhelmed by court prep? Want to summarize the latest end all-be all statistical research paper without having to solve a differential equation? Want to know the limitations of the assumptions of papers that hold themselves as authoritative in the discipline? Don't have time to attend that conference but want to stay informed?

Have no fear! AI is here....

Feel free to add to the thread.

Re: Fingerprints and the IAI (Inquisition of Artificial Intelligence) - A ChatGPT thread

Posted: Fri Mar 24, 2023 5:53 am
by Bill Schade
on the second report discussed (OIG report on the Mayfield Error)

#4 Lack of accountability

I believe that is a positive attribute since it looked at systemic causes instead of finding a scapegoat

For most of my years in the discipline (at least the first 25) the pattern evidence discipline said that any errors were the result of poor practitioner training and or ability. The "system" was perfect, the people making errors were bums and were run out of town.

The OIG report looked at what caused so many highly trained and respected examiners to make such a serious error

Re: Fingerprints and the IAI (Inquisition of Artificial Intelligence) - A ChatGPT thread

Posted: Sun Apr 02, 2023 1:22 pm
by Boyd Baumgartner
I'd have to disagree with that point RE: systemic error, the report says as much. For Mayfield anyway, the chatgpt response points 4 and 5 are mutually exclusive and I agree with it.

from the OIG Report, pg 209, the FBI LPU recognized three types of error:
  • Administrative errors
  • Systematic errors
  • Interpretive/Analytic errors
It then says:
The FBI Laboratory categorized the Mayfield fingerprint misidentification as an analytical/interpretive error, the most serious category of error.
Even the root cause of the error was put on the Examiner:
For this type of error, the Manual requires that the error be "discussed and/or documented with the Examiner to determine how and why the wrong conclusion was reached."

This is explicitly stating it's an examiner issue, and not a systemic issue.

To illustrate further, pg 96-97 the report states that the lab had the following systems in place:
  • ASCLD Accreditation
  • Training and certification programs
  • SOPs
  • Proficiency testing
  • Comparison methodology (SWGFAST)

what they deemed primary causes
  • Similarity of the prints - cognitive/perceptual issue
  • Bias from the known prints of mayfield (circular reasoning) - cognitive/perceptual issue
  • Faulty reliance on extremely tiny (Level 3) details - cognitive/perceptual issue
  • Inadequate explanations for differences in appearance - cognitive/perceptual issue
  • Failure to assess the poor quality of similarities - cognitive/perceptual issue
  • Failure to reexamine LFP 17 after the April 13 Negativo report - maybe systemic if it was a policy decision in writing somewhere
To put a bow on this point, what they deemed as corrective action was to send them back through the system that produced the error
pg 209/210 again
The Manual further states that four actions may be taken with respect to the responsible examiner: (1) immediate removal from conducting casework, (2) complete technical review of the examiner's past cases, (3) proficiency testing, and (4) training

Arguably with the exception of the last point of what they deemed primary causes, everything is put on the examiner and none of those explanations can shed light on the most important factor of the error. Namely, what was the SNP doing that allowed them to get it right? Certainly the most fruitful means of systematic review is a comparative analysis of two systems. I don't see where that was done.

While examiners may not get fired these days, it certainly doesn't lend itself to agencies taking accountability for systems that produce errors. It just results in examiners ending up on Brady lists which is basically just credibility assassination. I don't really see a difference. Agencies act in their best interest and that doesn't necessarily line up with Examiners' best interests.

That's why I like giving examiners the power of artificial intelligence for scrutiny of agency policies and research. Examiners answer for shoddy practices and research on the stand, not Managers.

Re: Fingerprints and the IAI (Inquisition of Artificial Intelligence) - A ChatGPT thread

Posted: Mon Apr 10, 2023 5:53 pm
by Boyd Baumgartner
I've had a couple people reach out to ask some questions. The short answer is, the applications are really only constrained by what you can think up. I found a Quality Manual online and a copy of the 17020 standards and asked it to evaluate the Manual against the Standards and spit out a detailed chart of where things may be off.


Re: Fingerprints and the IAI (Inquisition of Artificial Intelligence) - A ChatGPT thread

Posted: Mon Apr 24, 2023 11:18 am
by orrb
If you are not paying any attention to AI and AGI you better start now. Large Language Models (LLM) are going to change how we work dramatically. I have been learning and using these models for two months now and it is amazing. Two weeks ago I used it to help me write a federal grant proposal and two foundation grant proposals. Saved my about six days of work. We have AI attached to our SOP's. I use it to help write emails. Summarize research papers, online articles, etc.

This is fundamentally bigger than the internet and the iPone combined. Most of the models are free. The paid models are not very expensive. Get in and play with it. AI will not take your job, the person that knows AI is going to take your job. I had chatGPT answer the following:

How AI can assist a forensic scientist?

1. Automated image analysis for fingerprint identification
2. Improved DNA analysis techniques
3. Automatic detection of facial features for identification
4. Automated analysis of crime scene photos for evidence
5. Improved voice recognition software for analyzing audio recordings
6. Automatic analysis of handwriting for forgery detection
7. Detection of patterns in large sets of data to identify criminal activity trends
8. Improved predictive analysis for identifying potential criminal activity
9. Automated analysis of surveillance footage for suspect identification
10. Integration of AI into forensic science databases for more efficient data management and analysis.

I know that 1, 3, 4, 5, 6, 7, 8, 9, and 10 are already happening. Not sure where the DNA folks are. We can't hide from this wave of technology like some of use are still doing by not using digital image capture, digital documentation, digital data collection, etc. AI is here and it is moving fast. It is very exciting and I urge you to pay attention.