Outside the lab, artificial intelligence was first used to help researchers study specific patterns of linguistic and physiological characteristics between partners to predict conflicts in love.
Most of the conflict monitoring experiments conducted by couples were conducted under the control of psychological experiments. But in the Choice Mobile Sensing Project at the University of Southern California in Los Angeles, researchers took a different approach by collecting data on wearables and smartphones to study couples living in normal living conditions.
Their early experiments with 34 couples showed that the combination of wearables and machine learning-based artificial intelligence may make smartphone applications a future family relationship consultant. “In our current model, we can detect when conflicts occur, but we haven't predicted the signs before the conflicts,†said Adela TImmons, a clinical and quantitative psychologist at the University of Southern California (University of Southern California). Ph. D. Candidates. “Next, we want to be able to anticipate conflicts and send real-time tips, such as reminding couples to rest or do meditation exercises to see if they can prevent or reduce conflicts between husband and wife.â€
Predicting the contradiction between husband and wife in real life is not an easy task. However, if the machine learning algorithm can automatically identify the data pattern, which helps researchers to screen out the linguistic and physiological unconventional indicators of the partner, which can include representative indicators such as heart rate or skin reaction, it can more accurately identify the brewing conflict. sign.
At the Institute of Electrical and Electronics Engineers meeting, the University of Southern California research team detailed their approach. Before using an off-the-shelf machine learning algorithm to analyze the data, researchers must determine which key features they should focus on during the experiment to get the best predictors of conflict. Past psychological research has shown that conflicts between husband and wife are related to physiological indicators such as increased heart rate and skin conduction levels.
TImmons explained that couples tend to use specific language in conflicts, such as more person pronouns ("you"), more negative emotional words, and more certain words, such as "forever" or "from Do not". The 34 couples who participated in the one-day trial wore wearable devices, including wristband sensors that measure skin conduction, body temperature and physical activity. There are also sensors that measure heart rhythm. Each pair of partners also collects a recording of their conversations via a smartphone and allows them to be tracked by GPS.
To confirm that a conflict has indeed occurred, the smartphone will alert couples to report during the quarrel. Of the 34 couples, a total of 19 people reported conflicts in the experiment. Early experimental results are promising. This finding is consistent with the theory of love conflicts proposed in previous psychological studies. For example, the accuracy of negative emotions expressed in words associated with conflicts is 62.3%. When the machine learning algorithm analyzes all the data of many different indicators, the accuracy of detecting conflicts is as high as 79.3%.
TImmons said: "These models rely on machine learning." "Machine learning algorithms can perform classification experiments and perform accurate detection. Couple conflict detection requires a lot of data." Of course, 79.3% accuracy is far from actual application, still not Provide active counseling or similar interventions for couples.
Theodora Chaspari, a Ph.D. candidate at the University of Southern California's Signal Analysis and Interpretation Laboratory, and the project's researcher, said misidentification of conflicts could lead to unnecessary alerts. However, combined with the data of many different characteristics, higher accuracy seems to confirm that the mental state of the couple in conflict can be effectively inferred in many different ways.
Researchers are also faced with the challenge of collecting real data for couples, which is far more complex than data in the lab. They sometimes encounter missing pieces of data, such as couples turning off their smartphone audio recordings at specific times for privacy. Despite this, Chaspari hopes to collect more data from more couples to help machine learning algorithms eliminate these noises.
In the end, the University of Southern California team hopes to use their system to collect enough personal data to determine the conflict patterns of couples, and there is still a long way to go to improve system identification accuracy. Chaspari said: "We have a universal system now, but the challenge we face is how to make this system suitable for any particular couple." Accurately identifying conflicts can ultimately make the algorithm before the couple realize they are about to start fighting. Predict conflicts.
The next step in the University of Southern California team is to improve the accuracy of their current algorithms by collecting more data. For example, wearables and smartphone technology can help researchers gather more data about other factors, such as time spent on mobile phones, time spent on the Internet, and how much sunlight a couple receives during the day, such as subtle potential conflict predictions. For example, prolonged exposure to the sun can have an impact on personal emotions.
TImmons said: "Part of the premise to help these models work well is that there is a lot of data and a lot of behavioral characteristics." "In the next steps, we will add more conflict prediction factors."
72V Battery Pack ,Lithium Ion Battery Pack,Lithium Battery Pack,Battery Power Pack
Zhejiang Casnovo Materials Co., Ltd. , https://www.casnovo-new-energy.com