Crowdsourcing is better than ACE-V or CASE-V
Posted: Thu Nov 14, 2019 9:47 am
Did I just coin a new term, CASE-V for Case AFIS? I think so.
Two interesting papers. One recent and one a few years old both dealing with crowd sourced decision making.
Collective Intelligence in Fingerprint Analysis
Latent Fingerprint Value Prediction:Crowd-based Learning
Although, to be fair, Dror has published on a similar notion described as Distributed Cognition, but more showing a need for it due to bias of things like AFIS scores.
The Impact of Human–Technology Cooperation and Distributed Cognition in Forensic Science.
Crowdsourcing is just a way to minimize bias by focusing on the consensus of individual judgments thereby washing out the individual biases that may exist. We did something similar in the Show Me the Print thread.
It looks like crowdsourcing produces a 12% increase in accuracy. Contrast that with is what I believe to be a 10% increase from previously unidentified latents in the Case AFIS Paper. While the methods were obviously dissimilar, they both claim an increase in efficiency/accuracy.
Now we just need to crowdfund a crowdsourced AFIS, deploy it in the cloud, then we could all just work from home.....
Two interesting papers. One recent and one a few years old both dealing with crowd sourced decision making.
Collective Intelligence in Fingerprint Analysis
Latent Fingerprint Value Prediction:Crowd-based Learning
Although, to be fair, Dror has published on a similar notion described as Distributed Cognition, but more showing a need for it due to bias of things like AFIS scores.
The Impact of Human–Technology Cooperation and Distributed Cognition in Forensic Science.
Crowdsourcing is just a way to minimize bias by focusing on the consensus of individual judgments thereby washing out the individual biases that may exist. We did something similar in the Show Me the Print thread.
It looks like crowdsourcing produces a 12% increase in accuracy. Contrast that with is what I believe to be a 10% increase from previously unidentified latents in the Case AFIS Paper. While the methods were obviously dissimilar, they both claim an increase in efficiency/accuracy.
Now we just need to crowdfund a crowdsourced AFIS, deploy it in the cloud, then we could all just work from home.....
