You may have noticed chatbots becoming more commonplace in your life. Buying yourself a new sound system online? A chatbot pops up offering help. Need to confirm flight details? The airline’s chatbot can do that for you. Chatbots can offer a quick and easy way to get the information you need on the platform you are already using, whether that be a website, Facebook Messenger, or WhatsApp.

As they operate with low bandwidth and are simple to use, chatbots seem like they could be a beneficial technology to engage with recipients of our development and humanitarian programs, don’t they?

As with any technology, the use of chatbots is migrating from the private sector and becoming more prevalent in crisis and conflict response. However, documentation of lessons learned, successes and failures, and best practices of the use of chatbots in the development and humanitarian sectors are still few and far between. Through our work with Sesame Workshop, DAI conducted desk research and key informant interviews to examine use cases and review lessons learned for the use of chatbots in the international development and humanitarian sectors. In this blog, we summarize our key findings.

Image: Pixabay/Mohamed Hassan.

What Are Chatbots Used For?

From a review of chatbots for international development and humanitarian efforts, we found that use cases often fall into four main categories:

• Provision of information: Most chatbots fall into this category, creating an interactive web page or magazine-style format whereby users can ask questions or pick a topic, and receive text-based information, as well as information in other formats such as links, videos, or images. (Some examples include: Cosas de Mujeres, which provides information on gender-based violence; Voices of Venezuela, which provides information for migrants; and Girl Effect’s Big Sis, which provides young women with information on relationships and sexual and reproductive health).

• Training: This is also a common use, and often includes a gamified learning experience whereby users complete course materials through text-based content and quizzes (Digify Africa’s chatbots such as Kitso and Naledi; Learn.Ink, and Darsel)

• Broad support services: Comprehensive chatbots provide chat-based support and services such as reminders. These chatbots may provide healthcare services such as notifications on when to take medications and appointment reminders (Praekelt MomConnect).

• Reporting: Users can report issues such as instances of violence or misinformation to a chatbot. They may tag locations, provide links to misinformational websites, or send information about prevalent rumors. Users may even receive a response fact-checking the information. (Examples include UNAIDS Ask Marlo for COVID-19 mis/disinformation; Cofacts Taiwan for mis/disinformation).

What Is Chatbot Best Practice?

Across the different types of chatbots, varied settings, and different diaspora of users, we identified several lessons learned related to design, chatbot character or persona, and user experience. Below we outline these key lessons.

Good Design Is Key

Chatbot best practices are grounded first and foremost in user-centered design. Before the design of a chatbot, it is vital to understand the content the users desire, their usage of key platforms (Do they already use WhatsApp or Facebook?), and key barriers (Is data prohibitively expensive?). This information will help confirm whether a chatbot is the most appropriate path and, if so, inform the platform used and the details of design.

We all have our opinions on chatbots: Do you find them useful or frustrating? Our audiences are no different. Testing the idea of a chatbot, and then testing and gathering data throughout design is vital to understand whether this format will resonate with the target audience.

Chatbot Voice Should Be Accessible And Resonate With The Audience

The voice or persona of the chatbot is vital to encourage and maintain engagement. Firstly, the chatbot should speak in the language(s) spoken by the target audience and be aligned with its culture and nationality. Even when discussing technical topics, the chatbot should not use jargon or technical language that risks alienating people. Language used should be casual rather than formal: Remember, it is meant to be a conversation!

Evidence varies on whether a chatbot should have a defined persona—such as a name or character—to be effective and engaging. How effective a persona is, and the type of persona is fully dependent on the audience. Implementers should test options to see what works best.

Content Should Be Attractive and Relevant

Ultimately, users will come to the chatbot for their information needs. Content must therefore be timely and responsive to users’ needs, rather than pushing content the organization thinks people want. This sounds obvious, but getting the content right requires initial and regular user research to understand changing needs.

The User Experience Is Vital To Encourage Engagement

Users come for the content but stay for the experience. Chatbots are designed to simplify interactions and create seamless paths to content, so the chat flow should be intuitive and easy to navigate. Chatbots should replicate the flow of a real conversation: Think about how you would text a friend. As it is a conversation, users do not respond well to typing or reading lots of text. This is particularly important for those with low literacy. Where long text is necessary, break it up with headings, bullets, or different content types such as images, emojis, and memes.

Gamification Can Motivate Repeat Engagement

Gamification of the user experience can motivate repeat engagement. It can help break up the “routine” of usage, keep users intrigued by what the chatbot has to offer, and increase the enjoyment of the experience.

Rewards are a great way to motivate continued engagement with the bot, by providing positive reinforcement. This can be as simple as a gif to congratulate the user when they complete a journey in the chatbot, or access content for the first time. Quizzes are also a great way to encourage engagement in the app and break up the back-and-forth required to access content.

Push Notifications Should Be Used Cautiously

Perceptions of the benefit of push notifications are varied. For some, notifications are a good way to encourage a return to the chatbot. For others, push notifications are annoying and risk causing a user to disengage.

Most people will have used chatbots that send a generic chat when you don’t engage, such as “hey, come back” or “how are you?” These types of alerts are unlikely to draw in the user. Successful types of notification messages include motivational quotes, messages that promote new content, and messages that inform a user of their progress. Check your data: Notifications sent at times of peak usage are most likely to result in engagement.

The Future of Chatbots

As the use of chatbots in the international development and humanitarian sectors becomes more prevalent, we hope to see implementers share their failures, successes, and lessons learned. Working on a chatbot? We’d love to hear from you on how it’s going!

The jury is still out on how effective chatbots can be, but a few promising use cases give us hope that they can be used effectively for our sector. As always in digital development, it’s all about good design.