Personalized AI: Using Algorithms to Elucidate the Memory Recall Process
Updated: Aug 22, 2018
By Avery St. Sauveur
We’ve all been there: we’ve forgotten that important definition for an exam, forgotten the right way to that store you need to get to, forgotten that one friend’s birthday. We’re all subject to lapses in memory, as well as successes. Maybe we’ve successfully memorized the amino acids and their structures. We remember that one family vacation we took when we were young. We finally remembered where we put the car keys. We’ve all endured gains and losses in memory, but it’s when these losses outweigh the gains that things become difficult.
Researchers at the University of Pennsylvania (Ezzyat et al. 2017) gathered 25 different epilepsy patients with pre-implanted electrodes (put there for an unrelated study) placed both on the surface and within neuron-rich tissue inside the brain. These patients presented a convenient situation for the researchers; they could both record brain activity and initiate electrical stimuli in one system.
Each of the 25 participants was presented with lists of 12 words chosen at random. Each word in the list was quickly flashed on a screen for less than two seconds, and then a second-long blank screen bridged the gaps between words. After all 12 words were displayed, the participant would enter into a distraction period After this, researchers asked the participant to recall as many of the 12 words as possible in 30 seconds. They repeated the process with 25 different lists of words per session, and two or three sessions per participant. Throughout, researchers used the implanted electrodes to collect voltage measurements during the recall process. The result was a bank of personalized data for each of the 25 participants, used to create individualized algorithms. This artificial intelligence could predict the words the patient was likely to remember and which were more likely to be forgotten, simply by analyzing the patterns in voltage associated with both forgetting and remembering words.
These algorithms were used to directly impact memory outcomes. This not only allowed Ezzyat et al. to identify instances of memory failure, but it then allowed them to stimulate the region in which the failure occurred. The decision to stimulate was left up to the algorithm’s analysis; it would identify the moments of failure and then determine, based on the individual’s own brain activity patterns, whether or not stimulation would induce a gain in memory. The researchers found that stimulation of the lateral temporal cortex, a region of the human brain located on the side of the head and a hotspot for sensory input processing, produced the most promising results. With this site as the location for stimulation, Ezzyat et al. determined that the probability of item recall increased on average by 15% when compared to participants who did not receive lateral temporal cortex stimulation.
For years, dementia diseases such as Alzheimer’s have baffled scientists. Pfizer, after repeated disappointing clinical trials, recently laid off 300 employees, effectively ending its endeavors to develop drugs to target Alzheimer’s and Parkinson’s diseases (Dwyer 2018). We’re still in the dark about what actually causes it. And for the possible millions affected by dementia nationally and the millions more affected by those suffering, this is immensely frustrating. Researchers like Ezzyat et al. offer a glimpse into the valuable assistance that artificial intelligence can provide. While we aren’t there yet, maybe it’s possible to one day fully and safely embrace this human-computer connection to ward off the effects of dementia.
Ezzyat, Y., Wanda, P. A., Levy, D. F., Kadel, A., Aka, A., Pedisich, I., Sperling, M. R., Sharan, A. D.,Lega, B.C., Burks, A., Gross, R. E., Inman, C. S., Jobst, B. C., Gorenstein, M. A., Davis, K. A., Worrell, G. A., Kucewicz, M. T., Stein, J. M., Gorniak, R., Das, S. R., Rizzuto, D. S., & Kahana, M. J. (2017). Closed-loop stimulation of temporal cortex rescues functional networks and improves memory. Nature Communications, 1-8. doi:10.1038/s41467-017-02753-0
Dwyer, C. (2018, January 8). Pfizer halts research into Alzheimer’s and Parkinson’s treatments. National Public Radio. Retrieved from https://www.npr.org/sections/thetwo-way/2018/01/08/576443442/pfizer-halts-research-efforts-into-alzheimers-and-parkinsons-treatments
Artwork by Mykl Ambros