How to create a (almost) gender-friendly chatbot

In an increasingly digital world, businesses need fast and more efficient software. Clients demand real-time services and chatbots, a type of robot, seem to be a viable solution. Likely, our childhood fantasies pictured mighty robots whose sense of justice overcame the often weak human willingness to play by the rules, but we aren’t any close to that reality. Robots are an extension of ourselves, and carry the way we understand and organised the world. Currently, there is much noise out there about AI and how “we footprint” it with our social biases, gender included. In this article, we explore the possibility to create a gender unbiased chatbot.

Gossiping with the bots

In loose terms, a bot is software able to perform automated tasks. The bots that ‘live’ in chat platforms are chatbots. This type of bot can have a quasi-conversation with humans depending on the programmed answers to possible questions. Usually, users can ask questions or give instructions and bots respond or perform the action as requested.

Although this concept isn’t new, its popularity and usefulness grows every day as they have moved from simple textual format to artificial intelligence-powered systems.

How can chatbots be useful in business?

We find chatbots in several well-known platforms like Facebook, websites, mobile apps, etc., as creating and implementing a chatbot is becoming easier. They give businesses the possibility to reach larger audiences, increase engagement by improving on-site user experience, ease business operations and data collection, among other benefits. According to the Global Web Index, 75% of internet users are using chatbots for customer service or e-commerce.

Here are four advantages of using chatbots:

  1. It’s available 24/7

Have you ever noticed that some websites have a chatbot that pops up when you enter the site? These bots are slowly replacing contact forms, the need to send emails or even to place a call. Since they are virtual robots, they don’t get tired, they are always available and they respond to clients requests whenever needed.

  1. It increases client satisfaction and positive feedback

Chatbots don’t interact based on mood or emotions; they treat customers the way they were programmed: politely and patiently. They are able to provide a quick response to clients, satisfying the need for timely, faster services.

  1. It’s budget-friendly

Sticking to the budget is challenging especially when it involves people’s management and workforce needs. Chatbots are a convenient budget-friendly solution to team up with your client-focused departments and put in place a better user experience. They can process simple requests, answer questions and transmit important information to human agents.

  1. It can collect data more efficiently

Chatbot technology is powered by machine learning that helps collect data. It can easily monitor customer behaviour (i.e. buying patterns, preferences) and provide marketers with detailed, tailored information, for marketing strategies and content to be adapted to clients’ needs.

The use of chatbots has grown exponentially across major sectors. In fact, this solution offers more advantages than disadvantages (such as complex interfaces and the use of primitive algorithms). Nowadays, smart programs are part of our daily lives.

Chatbots and gender

Chatbots are not only changing the way we experience customer services, but also breaking gender stereotypes.

chatbots

In the past, most chatbots were female. Today, new bot start-ups are steering away from the typical female voice to adopt a more balanced mix between female and male voices. In many cases, they end up not choosing one or another.

The culture of a company and the developers and content programmers’ cultural background influence bots. The bot might be a robot, but its creator is a human being who unconsciously includes both virtues and biases in the system. It seems to be quite difficult to create chatbots that aren’t embedded with the unconscious gender stereotypes common in society.

If you had the chance to create a chatbot, how human would you want it to be? Dror Orenm co-founder of Kasisto, a company creating conversational AI platforms for the financial industry, says,

Bots can certainly seem very human-like as they possess more of our traits with the ability to understand, reason, and interact. That’s why the discussion itself is part of establishing the bot as a machine.

A bot needs a combination of behaviour analysis and machine learning to understand human requests, the client’s writing style for example, and know when to pass a more complex request over to a human.

Four recommendations for a gender-friendly chatbot

The excitement around artificial intelligence is understandable. A tangible example of how convenient it can be for businesses are chatbots. It’s up to us to decide what we want robots to do and to be in our world. Do we want a human-like being that makes the same mistakes we make? Do we want systems whose main capabilities are only linked to tasks automation?

Based on a rich exchange of ideas, we came up with four recommendations for a gender-friendly bot:

          • Be aware that a perfect unbiased chatbot is highly improbable

Having a completely unbiased bot is very difficult. Being aware that gender biases change depending on the culture is a first step to be responsible for one’s creations.

According to a study published in Science, AI picks racial and gender biases through written language. Without the proper supervision, a machine can quickly make gender associations between female names and family-related words more than career-related words, for example. Aylin Caliskan, post-PhD researcher at Princeton University and a team of experts tested the bias of common AI models. They matched the results with well-known psychological tests that measure human biases. According to her,

AI is biased because it reflects effects about culture and the world and language. […] So whenever you train a model on historical human data, you will end up inviting whatever that data carries, which might be biases or stereotypes as well. […] AI just captures the world as it is — it just happens that our world is full of bias.

          • Integrate gender unbiased data

A machine cannot recognise something it hasn’t seen before. The system can only be as diverse as the information introduced. (This research can help you understand this better.)

Being fully conscious of the different cultures, existing gender issues and biases, is already an important step to develop an inclusive, less biased program. Include your gender and diversity champion or an HR professional responsible for diversity in the software team. This team can be responsible for the analysis, selection and introduction of gender-friendly and inclusive language in any software development, including chatbots.

          • Train programmers and content creators in inclusive language

Changing gender stereotypes is a long-term goal but now is the time to act. This is particularly important when it comes to teams developing products and services for clients. We recommend you to coach chatbot creators on the importance of inclusive language. It can positively influence client relationships.

          • Teach AI-powered bots to filter information

As we mentioned before, robots learn what we want them to learn. At least, for now. Create and integrate algorithms in your chatbot to filter out potential gender-biased language.

What we think
Emilia Tantar, Former Artificial Intelligence expert at PwC Luxembourg

We’re entering a time when Artificial Intelligence is the new normal. It’s becoming an essential part of business strategies, especially in customer services. There are several advantages in using this type of technology. However, we need to be responsible for our creations and the information we provide to the systems. Even in a simple chatbot, we have to consider that reducing gender and colour biases is crucial and it can be achieved, partially, by language analysis and filtering. Ethical values should be included in the machine learning process to avoid negative views and stereotypes. Be responsible with your approach.

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to Top