Artificial Intelligence and Human Intelligence in Qualitative Data Analysis

Artificial Intelligence and Human Intelligence in Qualitative Data Analysis

António Pedro Costa, Universidade de Aveiro (Portugal). Investigador do Centro de Investigação em Didática e Tecnologia na Formação de Formadores (CIDTFF), Departamento de Educação e de Psicologia, da Universidade de Aveiro e colaborador do Laboratório de Inteligência Artificial e Ciência de Computadores (LIACC), da Faculdade de Engenharia da Universidade do Porto.

António Pedro Costa, University of Aveiro (Portugal)

Researcher at the Center for Research in Didactics and Technology in Trainer Training (CIDTFF), Department of Education and Psychology, University of Aveiro and collaborator at the Laboratory of Artificial Intelligence and Computer Science (LIACC), Faculty of Engineering, University of Porto.

When we approach a chatbot about the relationship between Artificial Intelligence and Qualitative Data Analysis, it mentions that it is essential to note that Artificial Intelligence cannot wholly replace human qualitative data analysis. The interpretation and understanding of qualitative data depend to a large extent on the context and experience (background) of the researcher. On the other hand, until the first tools to support data analysis emerged – the so-called CAQDAS (Computer-Assisted Qualitative Data Analysis Software) – researchers carried out much manual work, from the definition, coding, and recoding of categories, among others. Colours were used to more easily identify the units of analysis associated with a particular category, a resource transferred and still applied by current software packages. Their intellectual work was very focused on the inferential and interpretive aspects. However, they were limited by what could be performed manually. Most of the studies were “merely” descriptive, presenting as results the syntheses of the text units encoded in each category. Since this type of analysis focuses on the process, it allows the researcher to get involved with the data, extracting meaning from them, and connecting them with their theoretical framework. Collaboration encountered geographic barriers, and involving two researchers, for example, in data coding, was not always possible. When it was, comparing coded data was hard work and, one might say, almost impossible to carry out. In most of the procedures, which were quite time-consuming, the possibility of error was common. Transcribing interviews required the astuteness of the researcher, so the process would not take longer than expected.

Many researchers still do their data analysis without using any digital tools. Is it possible to do it? Yes, it is possible. Is it in the interest of the investigation and the researcher to run it manually? The current scenario almost compels us to resort to this type of solution. Tools have evolved and automated many procedures. Almost everything can and is being automated from transcriptions, encodings, and outputs. Is the researcher prepared to manage such an advance? As qualitative data analysis is oriented toward processes and procedures, to what extent will Artificial Intelligence add value to this area of knowledge? The dilemmas are identical to those of the 1980s when Cuilenburg, Kleinnijenhuis & Ridder (1988) already anticipated how Artificial Intelligence would benefit qualitative data analysis. These authors mentioned that the content analysis technique implied tedious tasks and that computers could facilitate some of them in the coding phase. Brent (1984) stated that qualitative data analysis involves a series of monotonous and time-consuming tasks, which most researchers would like to avoid. Even the most committed qualitative researcher quickly concludes that such procedures are not an economical use of their time. However, these tasks require sustained theoretical foundations and are challenging to delegate to third parties and, I deduce, to a software package.

CAQDAS are presented as innovative solutions to support qualitative data analysis. Users, for the most part, perceive that they do not need to carry out readings that allow them to establish their methodological, conceptual frameworks…the software will provide the necessary response.

AI is a branch of computer science that simulates intelligent behaviour in computers, from the premise that a machine can mimic intelligent human behaviour. In 1950, the British Alan Turing created a test to determine whether a device could exhibit intelligent behaviour indistinguishable from human behaviour. The Turing test is performed as follows: a human evaluator asks questions to two participants out of their sight, a human being and a machine. The objective is to tell the engine from the human, based on the answers to the questions. If the evaluator cannot distinguish the device from the human, the machine must have passed the test. The test became a significant milestone in the development of artificial intelligence and has been widely discussed and criticized ever since. Although the test has been questioned for its simplicity and limitations, it remains an essential benchmark in artificial intelligence.

There are several advantages to having a semantic web and incorporating automation processes and artificial intelligence in data encoding in qualitative analysis software packages (Costa & Minayo, 2018).

These authors stress that it can significantly increase the speed, rigour, and productivity of the analysis process, which can be especially beneficial when dealing with a large volume of data. Automated processes can quickly and accurately encode non-numeric and unstructured data, making analysis more efficient and reliable. This can save researchers time and resources, allowing them to focus on higher cognitive tasks such as data interpretation and contextualization. Furthermore, it can improve the quality and consistency of the coding process. “Human coders” can introduce biases and inconsistencies into the coding process, leading to errors and inaccuracies. Automation and artificial intelligence can eliminate these problems, providing more accurate and reliable results.

Incorporating automation and artificial intelligence into the analysis process can provide new perspectives and insights that may not be perceptible to “human coders”. However, it is essential to note that these automation and artificial intelligence processes are not intended to replace researchers but to complement and enhance their work. While automation can help with the initial coding process, “human” researchers are still needed to validate and audit the results. In addition, all stages of inference, synthesis, determination of implications, and contextualization will still depend on the human factor (Costa & Minayo, 2018). We can say, therefore, that the use of automation and artificial intelligence in qualitative data analysis should be seen as a tool to help researchers in their projects rather than a substitute for their expertise.

References

Brent, E. (1984). Qualitative computing: Approaches and issues. Qualitative Sociology, 7(1–2), 34–60. https://doi.org/10.1007/BF00987106

Costa, A. P., & Minayo, M. C. de S. (2018). O que podemos esperar da análise de dados qualitativos suportada por software? In M. A. Kalinke, M. A. V. Bicudo, & V. S. Kluth (Eds.), Atas do V Seminário Internacional de Pesquisa e Estudos Qualitativos. SE&PQ. https://sepq.org.br/eventos/vsipeq/documentos/P866236/50

Cuilenburg, J. J. van, Kleinnijenhuis, J., & Ridder, J. A. (1988). Artificial intelligence and content analysis – Problems of and strategies for computer text analysis. Quality and Quantity, 22(1), 65–97. https://doi.org/10.1007/BF00430638

Related Events

Related Publications

Share

Related News

A utilização de Inteligência Artificial no processo de escrita académica coloca-nos num equilíbrio delicado entre o potencial enriquecedor e a integridade.
Using Artificial Intelligence in the academic writing process places us in a delicate balance between enriching potential and integrity
The use of qualitative analysis in the metaverse world would enable food industries to gain a nuanced understanding of consumer behavior, preferences, and expectations.