Livin' on the MDedge

Artificial intelligence: The Netflix of cancer treatment


 

Chemotherapy, now streaming at just $15.99 a month!

It’s a lazy Sunday and you flip on Netflix, looking for something new to watch. There’s an almost-overwhelming number of shows out there, but right at the top of the recommended list is something that strikes your fancy right away. The algorithm behind the scenes is doing its job well, winnowing the universe of content right down to the few things you’ll find relevant, based on what you’ve watched and liked in the past.

rawpixel

Now, the almighty content algorithm is coming for something a little more useful than binge watching obscure 80s sitcoms: cancer treatment.

By plugging the fully sequenced genomes of nearly 10,000 patients with 33 different types of cancer into an algorithm powered by the same sort of artificial intelligence used by Netflix, researchers from London and San Diego found 21 common faults in the chromosomes of tumors, which they called copy number signatures. While cancer is a complex disease, when faults occur in those copy number signatures, the results were similar across the board. If X genetic defect occurs within a tumor, Y result will happen, even across cancer types. For example, tumors whose chromosomes had shattered and reformed had by far the worst disease outcomes.

The eventual hope is that, just as Netflix can predict what you’ll want to watch based on what you’ve already seen, oncologists will be able to predict the course of a cancer, based on the tumor’s early genetic traits, and get ahead of future genetic degradation to prevent the worst outcomes. A sort of “Oh, your tumor has enjoyed The Office. Might we suggest a treatment of 30 Rock” situation. Further research will be required to determine whether or not the cancer algorithm can get us part 2 of “Stranger Things 4” a week early.

Pay criminals, cut crime?

What is the best method for punishing those who commit wrongdoing? Fines? Jail time? Actually, no. A recent study says that financial compensation works best.

In other words, pay them for their actions. Really.

A gloved hand reaches out to grab cash wakila/Getty Images

Psychologist Tage S. Rai, PhD, of the University of California, San Diego, Rady School of Management, found that people who hurt others or commit crimes are actually doing it because they think it’s the right thing to do. The results of this study say play at the angle of their morality. When that’s compromised, the offender is less likely to do it again.

Four different experiments were conducted using an online economics game with nearly 1,500 participants. Dr. Rai found that providing a monetary bonus for inflicting a punishment on a third party within the game cut the participants’ willingness to do it again by 50%.

“People punish others to signal their own goodness and receiving compensation might make it seem as though they’re driven by greed rather than justice,” he said.

The big deterrent, though, was negative judgment from peers. People in the study were even more hesitant to inflict harm and gain a profit if they thought they were going to be judged for it.

So maybe the answer to cutting crime isn’t as simple as slapping on a fine. It’s slapping on shame and paying them for it.

Pages

Recommended Reading

Woman who faked medical degree practiced for 3 years
MDedge Internal Medicine
The Empire strikes out against one physician’s homemade star fighter
MDedge Internal Medicine
Nap length linked to cognitive changes
MDedge Internal Medicine
Fecal transfer could be the transplant of youth
MDedge Internal Medicine
TikTok challenge hits Taco Bell right in its ‘Stuft Nacho’
MDedge Internal Medicine
Common brain parasite linked to attractiveness, new study
MDedge Internal Medicine
Crohn’s disease research goes to the dogs
MDedge Internal Medicine
Today’s medical oxymoron: Healthy overconfidence
MDedge Internal Medicine
Parkinson’s disease could be hiding behind those nightmares
MDedge Internal Medicine
Microbiome’s new happy place: The beer gut
MDedge Internal Medicine