Summary: The investigators have shown that AI can see personality traits from a written text again, especially, now understand how these kinds make decisions. By using enchartable strategies such as combined gradients, the team found that certain names and language patterns provide for predictions based on major psychiatric structures.
Studies have found that five major factors are reliably accessible than the species of MBTI, as well as better agreement with ethical characteristics. This understanding can open a form of obvious personality, good mental test, HR, education and digital platform.
Key facts:
- Ai descriptive: Used by combined gradientents to express what influential words affect personality miscarriage, opening the “Black box” to make AI decisions.
- Big Five 5 VBT: The five largest model had reliable and mental allegations of the personality analysis based on AI rather than MBTI.
- Practical Use: The findings can enhance clinical, effective educational, HR processes, and AI variables.
Source: The University of Barcelona
The University of Reaches Group has shown that Intelligence Models can see how the personality traits from the texts, and the first time they are able to analyze how these programs make decisions.
These results, published in a journal One more plosOpen new ideas of understanding how ugly personality is also and that automatic inventory tools are transparent and reliable can be constructed.
This paper is signed three of UB experts: David Saetoos and David Gallardo-Pujol, researcher and neuroscience director, researcher in the Faculty of Mathematics and Computer Science.
To open a “black box” of algorithms
The study has analyzed the two advanced models in AI, Bert and Roberta, which processes the details of the text to find the helpful features of the two, responsible psychiatrics, sense of thinking, thinking and thinking.
“Psychology, there is a common humanity model and other subtle models, we use to understand and measure individual differences, feelings, and thinking”, researchers describing this both breadthalistic psychiatric.
The auditory documentation has been received from the two details of the information from both models (five hundred and MBTI), which were previously divided into humanity signs and types that included them.
Later, researchers have used selected techniques AI to look at AI models and see what language methods influence the identification of human characteristics in these documents.
“Description strategies allow us to 'open the black box' for algorithms, confirm that predicts based on the appropriate mental signals and not in the data”, call the authors.
Specifically, they have used a process called Gradients are combinedIt allows them to identify well which words or phrases that give up for predicting a specific human feature.
“This approach has made us logically and measure the significance of the different language items in the prediction of the model”, saying. For example, they have seen those words like hateTraditionally associated with negative aspects, it can arise in situations that are actually kind (“I hate seeing others suffering”).
“Without understanding how the model translates these words, we may receive wrong conclusions”, emphasizing.
This approach guarantees AI models, as it allows that models adapt to psychological views and provide patterns of genuine related languages with mental property.
MBTI model limitations
This study also promoted the limitations of the MBTI model compared to the Big Five, which showed a powerful basis for default personalized analysis and the classical psychometric analysis.
“Unless of being widely used in computer science and some used Psychology workers, the MBTI model has a great anti-humanity test, as our results indicate that they are most dependent on real patterns”, they see.
Default Face Request requests
The use of AI models can have a major impact on the psychological field.
“In these ways, psychologists will identify languages in line with various personality traits, in traditional ways, especially, especially,, are especially researchers.
In the medical field, the authors say they can help the examination “the first of the patients following the language or verbal.
They also said that they could play an important role in other areas: workers' choices, electronic choices, the development of high-quality assistants and Reformation, as it will help create the natural price and conversion.
“It is important to emphasize that all such applications should be based on sound scientific models and install the descriptive strategies, to ensure behavior and obvious use”, adding.
Despite skills, researchers believe that these kinds will not include traditional personality tests in a short time, but will use it and provide a negative and deep view.
“We see the appearance of the multimodal road, where traditional tests are included in natural language, digital behavior and other data sources to find a complete image of personality”, comment.
The related method will do, according to investigators, creates one to do each, providing 'a rich idea and added to one's personality'.
In this sense, AI models can be “very useful in the form of traditional data collection or where large information information requires properly”, adds.
To ensure research in some cases
The following steps in this study include analysis of other types of documents, platforms, languages and cultures to ensure whether the converted patterns are converted to separate. Studies want to test the use of these methods from another form of mind than personality, like the nations or attitudes.
The investigators work and compile multimodal data in this analytical data – such as voice-based or default text-related behaviors.
The team wants to 'work with the medical professionals and the Human Resources to assess the functioning of these actual arrangements for the world's arrangement, and ensure that they have a positive and moral impact.
About this AI and Ubuntu lesson
The author: Rosa Martínez
Source: The University of Barcelona
Contact: Rosa Martínez – University of Barcelona
Image: This picture is placed in neuroscience matters
Real Survey: Open access.
“The text speaks a lot: Understanding from personality from the nature of natural language” is David sateosoos et al. One more plos
Abstract
The text speaks a lot: Understanding from personality from the performance of natural languages
In recent years, advances in environmentalism performance (NLP) has enabled new forms of personality.
The article presents an unplanned investigation of AI strategies, especially gridients integrated, evaluating the decisions of the NLP decisions on human examination and ensuring their alignment and personal opinion.
We compare shopLogical performance (MBTI) and models of size (five hundred), using Essays and MBTI datasets.
Our way of our approach applies the Log-Odds with the former Dirichlet (IDP) and transformer based models (Bert and Roberta) to distinguish personality qualifications from the Scriptures.
Our results reflect on limited accuracy in human prediction, with NLP models that indicate the humanity signals in the text in accordance with previous lessons.
Our acquisition produces a united patterns through the use of languages associated with different personality traits, while highlighting important racism in the MBTI dataset that has placed such negative results.
The lesson emphasizes the NLP power to improving psychology and emphasizing the need for further study of full reading to monitor the skills of the technical technology.
