Mary Anderson gets a call at work on Thursday afternoon. It’s Peter; he says the Crawford’s signed the consent form and agreed to keep it quiet. Just before 5:00, she gets another call. It’s from Sam’s nephew, Jim. Jim gives her detailed instructions on how to access, input data, and run the AI-based algorithm remotely. It’s after her normal working time when she hangs up. She knows she has to do the search for Martha’s killer on her own time. She decided to start. First, she’ll do a simple case to verify she is using the algorithm properly. She runs the algorithm, without any of Martha’s data, using the State sex offender registry to see what the algorithm predicts for each of them for the time and place of Martha’s rape and murder. The algorithm predictions are what Sam said they might be. All of the probabilities are less than 1% except for Willie Sexton’s. It’s 4%, higher than any of the others, but not significant. Mary thinks, ‘With all his bumbling the sheriff did as good a job as possible given the information he had. Imagine how much effort he could have avoided if he had the use of this software.’ It’s nearly 9:00 when she finishes the first test and she’s hungry. She decides she’s done enough for the first day.
Friday after 5:00 Mary enters Martha Crawford’s data. She only has to input Martha’s name, address, schools attended, and job history. The AI program automatically matches her to the correct Martha Crawford on Facebook, Twitter, and Instagram, as well as to her bank accounts, her cell phone provider, all of her store credit cards, and loyalty cards. It even guesses correctly that she volunteered at Helping Hands and the gift shop. It didn’t guess at her being part of Doc Arra’s group. Mary is astonished but thinks, ‘At least it isn’t perfect. Maybe there’s still a role for us humans.’
Mary runs the algorithm again with Martha’s data. The result is the same. ‘I’m missing something,’ she thinks. She’s done enough for this night. She calls Pam. “Pam, this is Mary. Can you meet me for breakfast Saturday? I have run a couple of cases on the AI algorithm and the results aren’t helpful. I’d like your thoughts.”
Pam replies, “I have a better idea. Come out to my farm and have breakfast with me. We can talk freely here without worrying about people overhearing what we’re up to. Be here at 8:30, I’ll have coffee ready and breakfast on the stove.”
“Good idea. I’ll see you at 8:30. I’ll bring the printouts from the two cases I’ve run.” replies Mary.
Saturday morning after a quick breakfast Pam clears her antique dining room table while Mary spreads out the printouts she’s brought. When Pam sits down Mary says, “See, it’s just like Sam said it might be. All the probabilities are so low they are insignificant, even for Willie Sexton.”
Pam looks at the printouts and says, “Good to see the sheriff did as good as could be expected given the data he had.”
“That’s what I thought too, Pam. If he had this algorithm he could have avoided a lot of useless work,” replies Mary.
“I can see that, Mary. This algorithm has the potential to revolutionize police work even if it only eliminates useless work.”
“I think we’re missing something, Pam. That’s why I wanted you to see these results.”
Pam says, “It’s telling us none of the known sex offenders is likely the killer. How about trying robbers? You looked at the medical examiner’s report. Is there anything suggesting it was a robbery made to look like a sex crime?”
Mary replies, “According to the Crawford case file Sheriff Ericsson considered robbery and rejected it. I can run the AI algorithm with robbers easily. I’ll go do that before lunch. Meet me for lunch at the Harbor Deli. I’ll have the results to show you. No one will know what we’re looking at.”
“‘Good idea, see you at noon at the Deli, says Pam.
Before they order lunch Mary puts the printout for robbers on the table for Pam to see. Pam looks at it, nods her head indicating she understands, and says, “I see. Let’s order and eat. I’ll think about these results. Only two men with probabilities greater than 4%. Did you check their backgrounds?”
Mary says, “Yes, they are both local low-level thieves with no history of violence. We’re still missing something.”
“Yes, it’s not sex connected, but it has to be connected to violence. If you’re willing to do another run today try known murderers.”
Mary says, “I can do that. I’ll call you later this afternoon with the results.”
Pam is cooking her dinner when Mary calls. Mary says, “I did a run with known murders from the State criminal database. The probabilities weren’t even as high as for the robbers. We’re still missing something.”
“I recall hearing a rumor that Willie Sexton had a vendetta with Martha’s father. Maybe that’s what accounted for his probability being four times higher than any other sex offender. What if we try unknown murders who may have encountered Martha at some time? They could have a fear of being identified?”
Mary replies, “Yes, the Crawford file says Willie has a vendetta with Martha’s father. Your idea is interesting. I don’t know how to set that up. I’ll call Jim. Maybe he can set it up for me. It may be Monday before I can connect with Jim. Have a good Sunday.”
Mary reaches Jim during her lunch hour on Monday. She explains what she wants to run. Jim says, “Yes, it can be done. I’ll have to make some coding changes.”
He calls Mary back at 5:00 and says, “I made the changes and ran the algorithm for unknown murders that might have a connection to Martha. It turns up a 60% probability for a connection to a check-cashing store killer Martha tried to ID when she lived in Detroit. It didn’t ID the murderer.”
Mary says, “That’s good news, now we may be getting somewhere. Is there a way to adjust the algorithm to search for the check-cashing murderer?”
“I think so. Let me think about this. I am busy with another project I can’t delay. If I can think of how to set up the algorithm to search for that check-cashing store killer I’ll call you and tell you how to do it. It’ll be tomorrow.”
Mary calls Pam that evening and reports the progress. She tells Pam, “Now we know Martha had an encounter with a murderer. I'm getting hopeful now. Maybe Jim can tell me how to set up the algorithm to search for the identity of the murderer Martha encountered. I can hardly wait to hear from Jim.”
At noon on Tuesday Jim calls Mary. He tells her how to set up the algorithm to try another approach first. She’s to set it to look for the check-cashing store robbers in past data who may or may not have killed anyone.”
After 5:00 Mary makes the adjustments Jim explained. She runs the algorithm and finds a series of check-cashing store robberies in the metropolitan Detroit area. The first occurred ten years earlier. That robber was caught and served five years. There’s four more after he’s let out, the last one being the only one with murders. There’s an 85% probability the known robber did the other robberies. The algorithm reaches that conclusion based on the AI analysis of the known robber's social behavior and location data. It is late by the time Mary gets those results. She quits for the day. As soon as she gets home she calls Pam. She says, “Pam, I’ve got great news. The latest run shows an 85% probability a known check-cashing store robber is the same one Martha encountered. A few more runs with different parameters should give us a clearer picture. I’ll call again tomorrow night after I make more runs.”
Pam says, “This is getting exciting. Can I join you at 5:00? Maybe you can complete a run or two before we go to the museum for our Wednesday session.”
“Yes, that’s okay. I’ll take my lunch tomorrow. I can set up another run while I’m on my lunch hour.”
Mary is ready for the next run when Pam joins her at 5:00. Mary says, “I set up a run to ask the probability the known robber who murdered people on his last robbery; at least there’s an 85% probability he’s the same one, would kill Martha to avoid being identified and convicted of murder. Let’s run it and see what it says.”
Pam and Mary pace the room and chat waiting for the run to finish. It takes an hour. They hover over the computer screen when the run finishes. Mary says, “Look, it says there’s a 75% probability the known robber would kill to avoid being convicted of murder. I think we have our man if there’s any way he would have been in Pineville on the day Martha was killed. I can run the location cases tomorrow night.”
Pam asks, “Can it tell us the probability this man would rape his victim or make it look like rape to throw suspicion on a sex offender?”
Mary replies, “I’m sure Jim can tell me how to set that up. First, though, I want to run the locations. I know how to do that. Let’s go to the museum and tell Joe, Sam, and Peter what we have learned since last week. I’ll make the location run tomorrow night.”
Pam and Mary go to the museum where Joe, Peter, and Sam are waiting anxiously to hear what Mary has accomplished. She tells them step by step what she did and what results she got.
When she finishes Sam says, “Great job, Mary. That artificial intelligence-based algorithm is working like Jim hopes. If it connects that known robber and murderer to Pineville then he’s Martha’s killer.”
“I’ll know the probability he could have been in Pineville tomorrow night by 6:00,” says Mary.
Pam says, “With an 85% probability the known robber is the man who robbed the check-cashing store and killed two people I am surprised it’s only a 60% probability he’s the one Martha saw but couldn’t identify.”
“Jim says we shouldn’t expect probabilities to match like that, Pam. It’s because the AI algorithm may weigh social and psychological factors differently for different runs. What counts is the two probabilities are statistically significant, which they are,” Sam explains.
“Thank you for explaining that, Sam. I think that means he’s the same guy and he’d kill Martha to stop from being convicted of murder. I conclude he has a motive. Since the software isn’t perfect yet I think we have identified Martha’s killer even if the software doesn’t tie him to Pineville. Mary, can you look up this man’s background? Maybe one of us can tie him to Pineville if he’s been in the area recently.”
Mary replies, “Yes, I can read his police file while the location run is underway.”
Joe asks, “What are we going to do if we conclude we’ve identified Martha’s killer?
“Good question, Joe. I don’t think going to the sheriff is the right thing. He is liable to move too quickly and screw things up. How about this? Let’s meet at my farm tomorrow night at 6:30. I’ll have dinner ready for us and Mary will have the location run data and the known information on the suspect,” says Pam.
Sam says, “Yes, let's do that. I don’t want to wait until next Wednesday to plan our next step. Thanks for the dinner offer, Pam.”
Peter says, “Pam, would you mind if Sarah comes with me? She’s not as fearful as she was but I’d still not like to leave her alone at night.”
“Of course not, Peter. I’ll have plenty. My garden is near its peak. Be ready to eat a lot of veggies. Joe, you bring Eliza too.”
“I’ll bring dessert,” Sam offers.
“Sarah will love the opportunity to show off her bread-making skills. We’ll bring bread,” Peter says.
“I’ll be in trouble with Eliza if I don’t offer to bring something. How about we bring appetizers?” says Joe.
“You’re all making this easy for me,” Pam says. “Let’s go home. I’ll lock up tonight.”
No comments:
Post a Comment