Skip to main content
SearchLoginLogin or Signup

The Role of AI in Marketing: Gender Bias Problems in Thailand

The problem of bias in applying artificial intelligence in marketing can be tackled in a variety of ways. This chapter brings to light different case scenarios that illustrate the nature of input and design data, and suggests prevention strategies.

Published onApr 16, 2024
The Role of AI in Marketing: Gender Bias Problems in Thailand
·

The Role of AI in Marketing: Gender Bias Problems

Kamonwan Thongthep

Introduction

Nowadays, artificial intelligence technology plays a vital role in the business sector because of today’s higher competitive business environment, complex and dynamic marketplace, and limited resources, therefore, artificial intelligence technology has been used to support business activities. According to the research of McKinsey & Company's Global AI Survey: AI Proves Its Worth, But Few Scale Impacts published in PTT Express Solutions, it was found that global business adoption of artificial intelligence technology has increased by nearly 25% (year-on-year increase). Top executives in various organizations were interviewed and revealed that artificial intelligence can help businesses reduce costs more and generate more revenue for companies. 22% of those companies using artificial intelligence technologies have increased their revenues by more than 5 percent, and that revenue is likely to increase year over year. Artificial intelligence is most commonly used in Sale and Marketing Department because it can assist in customer recognition (sounds, faces, or objects), complex problem solving, language translation, or strategic planning, which can accurately help in business customer analysis. In addition, machine learning is a subfield of artificial intelligence that has more potential and the capability of a machine to imitate intelligent human behavior used to perform complex tasks in a way that is similar to how humans solve problems. It plays a role as a virtual assistant in processing big data and summarizing all data into an obvious picture of the targeted audience, which helps the marketing team design promotion activities faster and offers the right promotion to the right target group (Nipa Technology). Moreover, Mr. Dhanawat Suthumpun, Managing Director of Microsoft (Thailand) Co., Ltd. added that artificial intelligence played an important role in the marketing of the business sector in four dimensions; Engage or reaching the internal and external organization’s goals, Innovate or creating new business possibilities, Work or working style for superior efficiency and Solve or overcoming obstacles and social solving problems. Microsoft has applied artificial intelligence to predict future sales with high accuracy and suggested the most accurate and optimal method based on the amount of data generated in the past. From a study by the American Marketing Association in PTT Express Solutions, it was found that using artificial intelligence to communicate with customers leads to higher customer satisfaction and increased revenue. It also reduces the cost of hiring employees for large companies, meanwhile, it helps fill the gap for smaller companies that neither have the budget to hire a personalized customer service team nor use personalized marketing in company marketing strategy by using artificial intelligence to analyze customer insights to enable businesses to deliver the proper products, services, and promotions to suit the needs of the targeted customer group. From the case study of Subway, the sandwich brand has partnered with IBM Watson to bring artificial intelligence technology called WEATHERfx Footfall with Watson evaluating the correlation between weather, sales volume, and customer steps in each Subway store in order to tailor advertisement and offer promotions based on weather conditions. During hot weather, for example, artificial intelligence automatically pulls out the hot sandwich ads and then changes to the snack menu or cold drink ads instead. This results in more customers entering the store and greatly reduces wasted advertising costs. Therefore, using artificial intelligence in sales and marketing has become more and more popular and a part of an integral part of today's business sales and marketing strategies.

Although artificial intelligence used in customer analytics has many benefits for businesses, if it is designed with a bias, it leads to a negative impact on both the economic and social. The economic disadvantage is that the company must lose the opportunity to increase sales in some groups of customers because those customers are judged by artificial intelligence as low potential customers, not worth offering products and services. While the negative effect on society is those customers who are rejected by artificial intelligence cause them to lose their chances of being provided good service by companies. As often seen in examples in business articles, using artificial intelligence for the financial sector in the form of lending services, the results of AI analysis tend to favor male customer selection over female customers. This causes injustice and affects the company by losing the opportunity to reach new potential female customers. In addition, a bias in artificial intelligence creates problems for those who were denied credit, worsens quality of life, and may lead to other problems.

To the above-mentioned gender bias of artificial intelligence in business sectors, therefore, the researcher is interested in an in-depth study of such problems occurring around the world as well as studying the factors causing problems and guidelines of improvement for both public and private sectors. This research is expected to expand the knowledge and understanding of the problem and indicate improvement guidelines that policymakers can apply to cope with the mentioned problems.

Conceptual Framework

In this study, a conceptual framework was defined in two areas: the concept of artificial intelligence to demonstrate what artificial intelligence is being studied in the study and the conceptual framework of ​​applying artificial intelligence in marketing.

The concept of artificial intelligence is defined by the National Science and Technology Agency (2022), which described that an artificial intelligence system means software developed by using artificial intelligence techniques that can produce results such as prediction, recommendation, or decision which has an influence on the concerned environment. It must be working for the purposes defined by humans. Artificial intelligence techniques include:

  1. Machine Learning consists of supervised learning, uninstructed learning, unsupervised learning and reinforcement learning (reinforcement learning) by using a variety of techniques. including deep learning

  2. Techniques based on logic and knowledge consist of knowledge representation, inductive (logic) programming, knowledge base, deductive reasoning, and deductive reasoning. inference and deductive engine, symbolic reasoning, and expert system

  3. Statistical Techniques include Bayesian estimation and search and optimization methods.

  4. Robotics that are classified as intelligent robots. It has awareness, sensors, the ability to make decisions, own driver that uses artificial intelligence as its control or other techniques applied together with the cyber-physical system (cyber-physical system)

For the application of artificial intelligence in marketing, Ming-Hui Huang & Roland T. Rust (2021)’s the Marketing Research–Strategy–Action Cycle described the application of artificial intelligence in the marketing planning process covering various activities from the market research process to a marketing strategy creation and the actual marketing process, each step consists of many activities as shown in Figure 1.

รูปภาพประกอบด้วย ข้อความ คำอธิบายที่สร้างโดยอัตโนมัติ

Figure 1. Marketing Research–Strategy–Action Cycle

Source: Ming-Hui Huang and Roland T. Rust (2021)

From the above Ming-Hui Huang and Roland T. Rust (2021)’s Marketing Research–Strategy–Action Cycle, the researcher, therefore, used as a framework for the study. Artificial intelligence based on the conceptual framework of Ming-Hui Huang and Roland T. Rust (2021) described artificial intelligence as software that can deliver results such as forecasts, recommendations, decisions and exploring about how AI can be applied to marketing in all dimensions such as data collection for product design, customer analysis, a target customer group selection. This framework will be used for further study of biased artificial intelligence used in marketing.

Research Methodology

The following research process was conducted:

(1)     Finding preliminary information on the internet; for example, articles, research papers, and news related to the use of artificial intelligence in business sectors that encountered the experience of gender bias.

(2)     Data classification can be classified into 3 groups: problem, cause of the problem, and improvement guidelines.

(3)     Data synthesis and summarization 

Research Results

The results of the study can be divided into three parts. The first part is a case study of the underlying problem of bias in artificial intelligence that is applied to marketing in various types of businesses. The second part explores the cause of the problem that caused the bias. and the last section presents a preventative solution to the problem of such bias.

Underlying problem of bias in artificial intelligence

After the researcher studied the underlying problem of bias in artificial intelligence that is applied in marketing, it was found that the most frequent case studies were biased in financial services. A study by Yukun Zhang and Longsheng Zhou (2019) found that women's chances of obtaining financial services are somewhat lower than that of men due to some inferior attributes. For example, most women are paid lower than men, have lower skills, and have less stable work than men, so it was often viewed that female customers tend to have lower debt repayment ability than males. Many types of research in the past have found that women have lower access to credit than men. In addition, Ongena and Popov (2016) found that the possibility of accessing business credit is lower if the business owner is female because it is often considered that women own a lower ability to grow a business than men. According to The Print (2021), a research paper by Women's World Banking found that the credit gap between men and women in emerging countries is around $17 billion as a result of businesses losing the opportunity to reach a billion potential new customers.

Although the case study on the problem of biased artificial intelligence in marketing focuses on the financial services business, however, there are some studies that focus on the application of artificial intelligence in product design to meet customer needs. Eric Iversen (2019) explained that there is a gender bias in product designs in the past that pay in a favor of males' needs more than females, causing the produced design that are negative effects on women's use, such as seat belts which were designed mainly considering the male physical outlook but neglecting the needs of the female. This is because the designers and the samples used as the prototypes of the product designs are usually men, so it is a problem with the use of seat belts by women. It was often found that women were most likely to be injured in car accidents by seat belts. In line with Kat Ely (2015), the survey found that women involved in engineering and design have always been lower than men resulting in product development that does not meet the needs of women, for example, a sample group of research and product development for pharmaceutical products is usually men, although the response to the medicine by women is different from men. Therefore, when medicine was introduced to the market, the consequences of medicine used by women were found and caused problems for both society and business. These huge costs are all due to bias in the design process that focuses more on the participation of men than women. Another interesting case of product design is the design of the algorithms in climate control systems. The majority of the design team were men defining the cold weather protection system. As a result, most women feel that the air-con at work is too cold, while the weather is comfortable for men. Therefore, product design by humans or using artificial intelligence without the participation of women can lead to products with attributes that unmet the needs of women and the impact that may be too severe to be predicted.

Finding the cause of bias problems

From the above bias problems, the researchers searched for the causal factors from various articles and research papers and found that the main cause of bias is usually the input data and the design of data processing methods. A study by Sonja Kelly and Mehrdad Mirpourian (2021) on gender bias causes unfair discrimination in the financial services industry found that the cause is probably from the sampling process that does not cover some characteristics of the population (Sampling Bias). If most of the sampling was male then attribute characteristics data was taken from the male, therefore, the data used and trained to artificial intelligence is biased and causes problems. As it was explained in an above-mentioned case study of Algorithm Design in Climate Control Systems, input data from male sampling, so the climate control system was installed in the workplace, most women felt that the temperature in the workplace was too cold and that they could get sick (Kat Ely, 2015). It may be caused by a bias called Outcome Proxy Bias, which is a bias caused by the use of one type of data as a predictor of another data, for example, the reason why females have a lower chance of getting credit approval than males in financial services, this was partly viewed those women own less ability to pay off their debts than men because most women had lower incomes. Most of the people who were denied credit were women. In fact, not all women earn low incomes and have low debt repayment ability.

In addition, another type of bias that can occur is Labeling Bias, not very commonly found in marketing activities, however, one of the research projects by Sonja Kelly and Mehrdad Mirpourian explained a bias caused by one data representation of another, such as the profession of a doctor is male, the nursing profession is female. In fact, it is not always necessary for men or women to engage in that occupation. If this form of bias occurs in marketing activities, such as targeted analysis, it will result in people who engage in some types of occupations being ignored and losing the business opportunity to generate income from those types of customer groups.

Moreover, to these design process factors, Eric Iversen (2019) also presented another factor affecting gender-biased product design; that is, women's participation in studying and working in STEM (Science, Technology, Engineering, and Mathematics) fields is low. The study found the proportion of women working in design is only 20 percent, and it has remained a lower proportion than men from the past to the present. In addition, a survey of women entering STEM fields found that about 73 percent of them experienced negative work experiences. Therefore, it can be said that the low chance of women's participation in STEM fields is a major reason why the products designed and developed tend to be more in line with the needs of men than women and if product designers have applied artificial intelligence in the product design process, it would have the same outcome which is a prioritizing on male characteristic over female characteristics caused a biased AI.

Prevention and Mitigation of Bias

From studying the problem of bias, its causes, prevention and mitigation of biased artificial intelligence in marketing, it was found that preventing or solving such problems can be implemented throughout the artificial intelligence deployment process. It can be prevented the bias from the pre-processing before importing the data or In-processing or reducing post-processing bias as well as improving the organization's management at the same time and correcting the attitudes of people in the entire organization about giving opportunities to a diverse culture of participation. Considering the information mentioned above and additional research studies can be summarized as shown in Table 1.

Table 1 Possible solution to mitigate algorithmic bias

Dimension

Solutions

Pre-processing

Before data is used to train an algorithm, data scientists should use artificial intelligence to re-weight the data or assign different weights to examples based upon their categories of protected attributes and outcomes such that bias is removed from the training dataset. 

In-processing

In-processing, algorithms incorporate fairness into the machine learning training task itself. For example, an in-processing mitigation strategy might establish that women should be accepted at the same rate as men.

Post-processing

One technique which has gained popularity is reject Option-Based Classification by assuming that discrimination happens when models are least certain of a prediction. The technique exploits the “low confidence region” and rejects those predictions to reduce bias in the end game. This allows you to avoid making potentially problematic predictions.

Organizational adjustment

The entire organization should be involved, creating cross-disciplinary teams of data scientists and social scientists that identify, address, and mitigating bias.

National adjustment

Design effective government policies for supporting education and employment of women in STEM fields, more representation of women in designing and developing AI from the user interface, user experience, testing algorithms, and training can manage the gender bias.  

Source: Sonja Kelly and Mehrdad Mirpourian (2021) Ayesha Nadeem et al (2020) Eric Iversen (2019) and Josh Feast (2020)

Conclusion

Nowadays, artificial intelligence technology has played a pivotal role in the business sector, especially used in sales and marketing teams, because it helps in the matter of pattern recognition, complex problem-solving, language translations, and strategic planning resulting in more accurate customer analysis. While using artificial intelligence for customer analytics brings many benefits to businesses, if it is designed with a bias, it can have a negative impact on the economy and society. Therefore, the researcher is interested in an in-depth study of such a global problem. including studying the causing factors and implementing improvement guidelines in both the public and private sectors. It is expected that this research can help increase the understanding of the problem and indicate improvement guidelines that policymakers can apply to cope with the problem. In this study, the idea of artificial intelligence refers to software that can deliver results such as predictions, recommendations, or decisions, and the conceptual idea about how AI can be applied to marketing in every dimension, from collecting data to product design, customer analysis, and targeted customer selection, etc. This framework was used for the study of bias in applying AI to marketing work further. From the preliminary information, for example, articles, research, and news about businesses that have experienced using artificial intelligence with gender bias, it was frequently found that in most of the cases occurring is in financial services. Women's chances of obtaining financial services were lower than men's because of some inferior characteristics; for example, most women are paid lower than men, and most of the women have lower skills and less stable work than men. Therefore, it was considered that female customers tend to have lower debt repayment ability than males.

Furthermore, the study of using artificial intelligence in product design has addressed issues of gender bias such as seat belts, which were commonly designed for the male physical outlook and overlooked the needs of women's shape. It is because the sampling group used as the prototypes of the product designs and the designer team are usually men, that it finally caused problems for the use of seat belts by women.

One more interesting case in the product design field is algorithm design in climate control systems. Most of the design team are men who define certain attributes for cold weather protection systems. When climate control systems were installed in air conditioning at the workplace resulting in most women felt that the weather at work was too cold for women, while it was comfortable for men. There are several causes of bias, for example, it may be caused by a sampling process that does not cover certain characteristics of the population (Sampling Bias), for example, most of the samples taken from men make the attribute characteristics data caused by a bias called Outcome Proxy Bias, which is a bias caused by using one type of data as a predictor of another. The case of financial services is a good example; the reason why females have a lower chance of getting credit approval than males is a biased consideration that women are less able to pay off their debts than men because most of the women have lower incomes.

Another type of bias that can occur is Labeling Bias, the perception that the profession of a doctor is male while the nursing profession is female, although it is not always necessary for men or women to engage in that occupation in reality. If this form of bias occurs in marketing activities such as analyzing target customer groups. It will result in people who engage in certain occupations being dismissed and cause the business to lose an opportunity to generate income from that targeted group. Apart from that factor in the aforementioned design process, the low participation of women in studying and working in STEM (Science, Technology, Engineering, and Mathematics) fields is another reason why product development tends to meet the needs of men more than women. The problem of bias in applying artificial intelligence in marketing can be tackled in a variety of ways which can be prevented before importing the data, in-processing, or post-data processing, including improving the organization's management system at the same time. Moreover, improvements at the level of governmental policies in encouraging more women's participation in STEM education and careers are suggested. 

References

Ely, K. (2015). The World is Designed for Men. Source: https://medium.com/hh-design/the-world-is-designed-for-men-d06640654491

Feast, J. (2020). Root Out Bias at Every Stage of Your AI-Development Process. Source: https://hbr.org/2020/10/root-out-bias-at-every-stage-of-your-ai-development-process.

Huan, M-H and Rust R. T. (2021). A strategic framework for artificial intelligence in marketing. Source: https://link.springer.com/article/10.1007/s11747-020-00749-9

Iversen, E. (2019). Left Out by Design, or How the STEM Gender Gap Leads to Lousy Products. Source: https://start-engineering.com/start-engineering-now/2019/4/25/left-out-by-design-or-how-the-stem-gender-gap-leads-to-lousy-products

Kelly, S. and Mirpourian, M. (2021). Algorithmic Bias, Financial Inclusion, and Gender. Source: https://www.womensworldbanking.org/wp-content/uploads/2021/02/2021_Algorithmic_Bias_Report.pdf

Nadeem, A., et al. (2020). Gender Bias in AI: A Review of Contributing Factors and Mitigating Strategies. Source: https://aisel.aisnet.org/cgi/viewcontent.cgi?article=1048&context=acis2020

National Science and Technology Development Agency. 2022. ประกาศสํานักงานพัฒนาวิทยาศาสตร์และเทคโนโลยีแห่งชาติ เรื่อง แนวปฏิบัติจริยธรรมด้านปัญญาประดิษฐ์. Source: https://waa.inter.nstda.or.th/stks/pub/2022/20220331-ori-ai-research-integrity-guideline.pdf

Nipa Technology. DATA & Technology. Source: https://nipa.co.th/th/solution/ai-marketing Business Today. 2019. Microsoft มอง “ปัญญาประดิษฐ์ คือ การประดิษฐ์ปัญญาคน”. Source: https://www.businesstoday.co/bt-news/12/11/2019/11944/

Ongena and Popov. (2016). Gender Bias and Credit Access. Source: https://www.researchgate.net/publication/312068282_Gender_Bias_and_Credit_Access

PTT Express Solutions. 5 บทบาทสำคัญของเทคโนโลยี AI ในโลกธุรกิจ. Source: https://blog.pttexpresso.com/ai-technology-business/

The print. (2021). Biased AI systems contributing to $17bn gender credit gap in emerging markets: Study. Source: https://theprint.in/economy/biased-ai-systems-contributing-to-17bn-gender-credit-gap-in-emerging-markets-study/639557/

Zhang, Y. and Zhou, L. (2019). Fairness Assessment for Artificial Intelligence in Financial Industry. Source: https://arxiv.org/pdf/1912.07211.pdf

Comments
0
comment
No comments here
Why not start the discussion?