In AI — Feb 20, 2024

Creating Inclusive Conversational Interfaces: Challenges and Solutions

By: Renata Dzotova 6 min read

Introduction: The Rise of Conversational Interfaces and the Need for Inclusivity

In a world where conversational interfaces shape most of our digital experiences, inclusivity is no longer a distant goal but a pressing necessity.

Generally speaking, inclusivity is about putting people first and making them feel welcomed through delivering content that is clear and accessible to everyone. I believe, inclusive design is a way to show respect to all humans, regardless of their physical, cognitive, cultural, or situational characteristics. And as designers, it’s our ethical responsibility to make products that work for all.

As for conversational interfaces, they play a crucial role in promoting inclusivity by providing alternative means of interaction. Voice assistants, for example, enable people with visual impairments or physical disabilities to navigate through their devices using voice commands. Whereas, AI chatbots offer support and assistance to users who struggle with traditional UI or prefer a more conversational approach.

Like with any design, creating an inclusive product is a process that takes time and multiple iterations until it gets there. To help you build an inclusive conversational interface, let’s discuss the common challenges that might arise along the way and explore existing strategies to overcome them.

Overcoming Challenges in the Creation of Inclusive Conversational Interfaces: A Tactical Approach

1. Lack of diverse training data for conversational AI models: addressing biases and ensuring representation.

We all see the world differently thanks to our unique human characteristics which tend to evolve over time. Our perception of the world is influenced by such factors as age, gender identity, nationality, culture, physical and cognitive abilities, sexuality, language, permanent / temporary / situational disabilities, political opinion, education… This list could go on forever.

When it comes to conversational AI models, one of the common challenges is the lack of diverse training data. It’s natural for us, humans, to hold biases and stereotypes and, often, it happens unconsciously. The goal of inclusive design is to reveal, acknowledge and eliminate any biases which may arise from various sources, such as biased training data or biased human interactions used for model development. Addressing it is essential to ensure that conversational AI models are not stereotypical in their responses.

To tackle this problem, make sure to involve a diverse range of training data that includes input from people with different backgrounds, cultures, languages, and perspectives. This diversity helps in mitigating bias by exposing the model to a broader spectrum of ideas and experiences.

Ensuring representation in conversational AI models involves ensuring that different demographic groups are adequately represented during the development process. This includes involving individuals from different ethnicities, genders, ages, abilities, and socio-economic backgrounds. If your company is developing conversational AI models make sure you:

Actively seek out diverse datasets for training purposes.
Consider partnering with organizations or communities that can provide insights into underrepresented groups' experiences.
Continuously monitor and evaluate after deployment to identify any biases or imbalances present in the system's responses.
Incorporate regular updates based on user feedback to help improve accuracy and reduce any potential bias over time.

Helpful resource to research inclusive design practices - Inclusive Design Methodology by Microsoft

Social and cultural sensitivity is indeed a challenge when it comes to conversational interfaces. However, advancements in natural language processing (NLP) have provided a solution to better understand different languages and speech patterns.

NLP models are designed to analyze and interpret human language. With the help of these advanced models, for example, copywriting assistants, can now better comprehend the nuances of various languages and dialects. This enables them to effectively communicate with users from diverse cultural backgrounds. Thanks to NLP advancements, copywriting assistants can take into account the specific customs and traditions associated with different cultures and, as a result, provide more relevant and culturally sensitive responses.

In addition, NLP models assist in understanding different speech patterns that may be specific to certain dialects or accents. By recognizing these variations in pronunciation or grammar usage, copywriting assistants can adapt their responses accordingly for improved communication with users.

3. Socioeconomic and digital divide: optimizing conversational interfaces for areas around the world that do not have access to high-speed internet.

The socioeconomic and digital divide refers to the unequal access to digital technologies and the internet among different populations, mainly because of disparities in income, education, and infrastructure. While some areas of the world have the necessary infrastructure for high-speed internet, we should optimize conversational interfaces with a global perspective in mind. When designing conversational interfaces for underprivileged areas, you should consider the following:

Ensure that your conversational interface supports multiple languages. Catering to diverse user populations will help bridge language barriers and improve accessibility.
Keep in mind that users from different socioeconomic backgrounds may have varying levels of digital literacy. Design a user-friendly interface that is easy for everyone to navigate, regardless of their technical expertise.
For regions where internet speed is limited or unreliable, prioritize optimizing your application's performance by reducing bandwidth requirements and minimizing data usage.
Enable offline functionality whenever possible to let users access certain features even when they are not connected to the internet.
Tailor content within your conversational interface based on cultural sensitivities and local contexts specific to each region you target.
Collaborate with local organizations or communities who understand the needs of specific regions better than anyone else. They can provide valuable insights into optimizing your interfaces according to local requirements.

4. Accommodating users with disabilities or special needs: catering to diverse user needs and accessibility requirements.

Accommodating users with disabilities or special needs is another essential aspect of creating an inclusive digital experience. By catering to diverse user accessibility requirements, we ensure that everyone has equal access to information and services we offer.

There are various ways in which we can accommodate users with disabilities or special needs, such as:

Provide alternative text for images allows visually impaired users to understand the content of the image through screen readers.
Use clear and concise language to help individuals with cognitive disabilities or learning difficulties comprehend the information more easily.
Ensure a logical structure by using headings, subheadings, and bullet points to make it easier for screen readers to navigate through the page.
Implement keyboard navigation for users who have difficulty using a mouse and ensure that all interactive elements can be accessed using the Tab key.
Use sufficient color contrast between text and background to improve readability for individuals with visual impairments.
Provide captions or add transcripts for videos and audio files to enable individuals who are deaf or hard of hearing to access the information presented.
Offer options to resize text to accommodate users with visual impairments who may require larger fonts for better readability.
Conduct regular testing with assistive technologies such as screen readers to ensure that your product is compatible across different devices and platforms.

Make sure to do more research on other internationally accepted accessibility requirements WCAG (Web Content Accessibility Guidelines) and consider those in your design.

Why Are Companies So Interested In Conversational Interfaces?

Creating inclusive conversational interfaces comes with several benefits.

Firstly, inclusive design obviously leads to increased user engagement and satisfaction. When users feel that an interface accommodates their individual needs, they are more likely to engage with it actively which leads to higher retention rates and increased usage of the interface.

Secondly, inclusivity enables a broader reach by taking into account diverse user needs and preferences, which makes applications more accessible to various demographics and thereby opens up new markets and audiences for businesses.

And finally, creating an inclusive conversational interface enhances brand reputation. Businesses that prioritize inclusivity demonstrate their commitment towards diversity and accessibility in their products or services. This not only helps in building trust among users but also contributes positively towards the overall brand image.

Conversational AI interface is the future. Over the next several years, it will significantly enhance many aspects of our lives. By designing inclusive user experience and making artificial intelligence accessible for everyone, we will eliminate social barriers and provide equal opportunities for individuals of all backgrounds and abilities around the world. And we, as designers, are the crucial contributors to this transformation.

References

Inclusion. Apple Developer Documentation. (n.d.). https://developer.apple.com/design/human-interface-guidelines/inclusion

4 steps to inclusive conversational UI. CIO. (2021, October 18). https://www.cio.com/article/189401/4-steps-to-inclusive-conversational-ui.html

Microsoft Inclusive Design. Go to Microsoft’s Inclusive Design website. (n.d.). https://inclusive.microsoft.design/

Admin. (2023, June 12). Reducing social discrimination and promoting inclusion with NLP. LMLM. https://lean-mean-learning-machine.com/machine-learning/natural-language-processing/reducing-social-discrimination-and-promoting-inclusion-with-nlp/#:~:text=NLP is also promoting diversity,well as reducing social discrimination.

Joyce, A. (2023, August 10). Inclusive Design. Nielsen Norman Group. https://www.nngroup.com/articles/inclusive-design/

University, S. S. B. C., Sarcar, S., University, B. C., Waterloo, C. M. U. of, Munteanu, C., Waterloo, U. of, Toronto, J. S. U. of, Sin, J., Toronto, U. of, Toronto, C. W. U. of, Wei, C., Lleida, S. S. U. de, Sayago, S., Lleida, U. de, Technology, E. U. of, Insight, B., University, R., Google, & Metrics, O. M. A. (2023, July 1). Designing conversational user interfaces for older adults: Proceedings of the 5th International Conference on Conversational user interfaces. ACM Conferences. https://dl.acm.org/doi/10.1145/3571884.3597438