Making the World Better, One Artificial Conversation at a Time: How Chatbots Are Powering the Social Impact Sector
Maria Dyshel19:172021
Chatbots are not only for brands; they can reach girls at risk of trafficking
About 35,000 Nepali girls and women are trafficked to India every year, as the speaker cites, and the safety information meant to protect them is, in her words, boring. Maria Dyshel, Forbes 30 Under 30 social entrepreneur and CEO of Tangible AI, has helped build more than 20 chatbots and shows how nonprofits use conversation to reach people other channels miss. So what separates a chatbot that truly helps from one that merely exists?
See how chatbots engage, support and serve people, plus the do's and don'ts of building one
Maria Dyshel, CEO of Tangible AI, argues that chatbots are not only commercial tools but can engage, support and serve the constituents of nonprofits and social businesses. Drawing on bots her team built, including Maya for anti-trafficking education with Plan International and Cindy for imposter syndrome, she explains why conversational AI works and what it takes to build a good chatbot rather than just any chatbot.
“Cause let's admit it safety is boring. No one likes to talk about safety.”
“And then finally here, it's not necessarily easy to build, it's easy to build a chatbot, but it's not easy to build a good one.”
What you will learn
Explain why messaging reach, texting habits, impatience and our wiring for conversation made chatbots take off.
Distinguish three ways chatbots help social impact organizations: engaging, supporting and serving constituents.
Use interactivity such as quizzes and choose-your-own-adventure stories to make dry content memorable.
Plan a chatbot project from user research through conversational design, iterative build, pilot and scaling.
Avoid common mistakes such as overreaching on scope and leaving a launched bot without ongoing correction.
Key takeaways from the speaker
How nonprofits and social businesses leverage chatbots to scale their impact.
Real-world use cases of using conversational solutions for good
Do's and Dont's of creating a pro-social chatbot
Details
Level
Beginner
Format
Talk
Technical depth
Introductory technical
For
All career stages
Length
19:17
Published
June 2021
Questions this talk answers
What is conversational AI?
It is technology that lets people and computers communicate the way humans do, through natural back-and-forth conversation. It works by text through chatbots or by voice through assistants such as Alexa, Google Home or Siri.
Why have chatbots spread so quickly in recent years?
As the speaker cited, messaging platforms reach about 5 billion users a month, even in places without electricity or running water, younger people prefer texting, and people expect instant answers. Humans are also wired for conversation and tend to treat chatbots as human.
How do nonprofits and social businesses actually use chatbots?
They use them to engage (Maya teaches Nepali youth about safe migration through quizzes and stories), to support (Cindy helps people reframe imposter thoughts) and to serve (step-by-step guidance on COVID-19 rules for small businesses).
Why is iteration more important for chatbots than for other digital tools?
Chatbots give users far more freedom than graphical interfaces, so people can type anything from a curse to an unexpected query. Getting the bot in front of users repeatedly and improving it based on real interactions is therefore crucial.
What are the biggest mistakes when building a pro-social chatbot?
Trying to do too much, such as expecting to build a Siri in a few days, instead of doing one job really well. And leaving the bot alone after launch, when it needs constant correction as it meets new situations.
About this talk
From customer service to messenger marketing, chatbots are now a part of the digital landscape. But can Conversational AI be used only for commercial purposes? Recently, more and more social impact organizations worldwide embrace chatbot technology to reach more constituents, educate the public and improve their operations. Around the world, chatbots are helping people to apply for financial aid, educate teens about sexual health, and provide information and support in humanitarian settings. In this session, we'll explore how non-profits and social businesses use conversational solutions to scale their impact, and share lessons we learned from developing pro-social chatbots for a better world.
Making anti-trafficking safety content engaging (73 s)
Chatbot do's and don'ts (54 s)
And thank you for joining this session. And I think we're ready to kick it off. So hi, everyone. Welcome to women tech. I'm super excited to be with you here. And today we are going to talk about changing the world, one artificial conversation at the time, basically how chatbots and conversational AI are helping social impact. organization to do their important work. So before we dive into that important topic, what I want to take you through is my personal journey of how I came to be at this intersection of this emerging new technology of conversational AI and social impact.
So my story starts as, you know, a lot of professional stories. I was a computer science student. But since my very early days, since my first university days, I was always socially active. So while I was doing my bachelor's and master's in Hebrew University in Jerusalem, and then developing algorithms and mobile and Google, and then managing AI project in Sanofi.
Speaker 1: I was always doing something on the side. I was running an environmental campaigns. I was building technology for underserved populations with engineers without borders. Speaker 1: I could found in a nonprofit and I ended up traveling to India and helping a local organization help rural nurses with a mobile application. Speaker 1: So I was living this double life for more than a decade, actually. And it always bugged me. Isn't there a way to do what I love?
Speaker 1: For the causes that I find important. Isn't there a way to take something that I'm good at and use it to improve the lives of those that were less lucky than I am? Speaker 1: And this is why about two years ago, I took the leap and Tangible AI came to be. Speaker 1: So Tangible AI started with a simple mission, how we can put artificial intelligence, this powerful tool that's transforming the whole industries into those into the hands of those that solve humanity's most pressing problems.
Speaker 1: And, you know, this was the original idea. But one of the beautiful things about humans is that we collaborate. And as I met my co-founder, Hobson Lane and my wonderful team, I realized that you can't just address just one part of it. Speaker 1: That the solution, the way to make AI prosocial and serving the prosocial needs needs to be more holistic and complex.
Speaker 1: So today, Tangible AI is, in fact, three things. We still do a lot of consulting. We help nonprofits and social businesses with custom solutions and AI strategy. Speaker 1: We also build our own solutions inside our innovation hub. And finally, we are on an internship program that allows us to create a new generation of AI professionals that understand Speaker 1: The basics of what it means to be prosocial and how to create an AI that's a prosocial force.
Speaker 1: So we work in variety of technologies. We help for a lot of our customers. What we do is we help them just understand how AI can be good for them and for their business and for their nonprofit. Speaker 1: We build custom solutions using data science and machine learning. And a big part of what we do are chatbots and conversational AI.
Speaker 1: So, um, well, this is what we are going to focus on today. And before we dive into, you know, what actually happened, how organizations actually use conversational AI, let's just give a brief definition of what it actually is. Speaker 1: So to make it simple, conversational AI is this technology that allows people and computers to communicate with each other, just as humans communicate between themselves.
Speaker 1: Using human language and natural back and forth interaction that we usually call conversation. It can usually be done in two ways. Either it's done through text where you basically message with an application called chat bot. Speaker 1: And the same way you would message with a friend or with a family member, or it can be done by voice when you actually talk with your, with your virtual assistants through a smart speakers such as Alexa or Google home or through an app like Siri.
Speaker 1: And one of the things that we saw in the last, I would say five or six days is really an explosion of chatbots. Speaker 1: In the first chatbot was created somewhere in the sixties, but until five years ago, there were mostly a gimmick, something that people would play with something that would show how far we got with AI.
But lately we've, we've seen chatbots almost everywhere. And there are a few reasons for that. Speaker 1: One of them, and probably the biggest one is that currently there are as much as 5 billion users on messaging platforms every single month. That's the biggest playground and the biggest audience you can imagine. It's to the level that some, in some underdeveloped rural areas, people won't have access to electricity or running water.
Speaker 1: But they will have access to phones, even simple phones, and they can message another important factor is that the younger generations increasingly turn to texting as a means of communication. So most of us would prefer texting rather than picking up the phone and making a phone conversation. Speaker 1: And finally, as the digital technology is penetrating deeper into all the areas of our lives, we're also becoming more impatient.
So more and more, we expect the right answers, the right informations right there immediately at the tip of our fingers. So all the combinations of all of that calls for the solution. Speaker 1: And then one of the reasons why chatbots came in and solve all these needs is just our human nature. So we are wired as humans to communicate through conversations from the ancient times, the medieval times, to the modern times.
A lot of things changed for us, the way we commute, the way we live, the way we cook. Speaker 1: A lot of things are really different, but we still exchange information and ideas. We still bond and organize using human language and using this interactive turn-based thing that we call human conversation. And this is why it's so powerful.
Speaker 1: One more thing that really served chatbots in the recent years is a phenomenon called pareidolia. Speaker 1: And what is pareidolia, this, you know, scary, big word, just look for a second at this image and tell me what you see. And I'll actually look at the chat to make sure that you still hear me. Yeah.
So I see that people are seeing a smile. Speaker 1: People are seeing, people are seeing a basically a humanoid face. And the reason why it's happening is because as humans, we are wired to see human faces and human shapes or humanoid shapes everywhere. We see them in the clouds. We see them on the surface of Mars.
We saw it everywhere because this is how our brain discovers patterns in the surrounding world. So when we encounter something like a chatbot, something that, as you say, walks like a human and talks like a human, our brain thinks that must be human. And we make this additional step of personification and we start treating chatbots as human.
Speaker 1: Now, it's for sure, it's a big challenge. It's for designers that need to design those conversations, but it's also a beautiful opportunity to take this most powerful thing, which is human conversation and use technology to scale it and help it and help billions of people around the world to have access to better information and better support.
So there is no wonder that with all this being said, the amount of chatbots have just exploded in the last five or six years on Facebook alone. There are about half a million chatbots and it's growing every day. So we have chatbots in customer service, in sales, marketing, in operating our team, in operating our DevOps, in managing HR.
Speaker 1: We have bots that support us during mental health, we have educational chatbots, there are really a variety of chatbot applications out there. And in my recent role in pharma, I deal with a lot of these sales and customer service bots, but it always bugged me. Can we take this technology and can we really make it serve the underserved people that need more support than others, people that need more help than others.
Speaker 1: And today we're going to, to look and see that actually chatbots are not just for the commercial sector and they actually help quite a lot. A lot of organizations that are actually trying to make the world and our planet better. So we're talking about the, some of the very important functions that chatbots perform in helping social impact organizations.
Speaker 1: And this is engaging, supporting and serving these organizations constituents. We're going to look at an example for, for each case study. And, but there are many more examples that you'll be able to explore on your own. So first let's start with engage. So the challenge that we're dealing with is non trivial and acute. Speaker 1: About 35,000 Nepali girls and women between five and 18 are traffic from Nepal to India every year.
Now it's a horrible number and it sounds horrendous, especially to us in the West. But for teenage women and girls and boys living in Nepali villages. Speaker 1: This is just life, right? They're just living their life and they're running into these difficult situations that they don't know how to handle. And it's a difficult conversation to talk about.
Cause let's admit it safety is boring. No one likes to talk about safety. Think about what was the last time you were at the safety booklet. Was it something that excited you? Was it something that engaged you? Speaker 1: So the challenge is how we take this boring subject of, of keeping yourself from trafficking and make it accessible.
So what we did with an organization called plan international is that we build a child, but called Maya. Which what it does, it takes these topics of keeping yourself safe and safe migration for work. And it makes them interactive. Speaker 1: So all the informational parts inside Maya are interactive as are, and it also has quizzes and choose your own adventure stories.
So it really makes this content not only accessible, but also memorable. And nonprofits have realized quite a while ago that this factor of interactivity is something that really can help you get the message through. Speaker 1: So today there is a variety of solutions that are built by nonprofits everywhere that can help you empathize with a person with learning disability or educate you about, about your body and how to keep yourself healthy, or even talk to you about gender bias in AI.
So that's the first use case, how we engage and educate people with messages that we want to be remembered. Speaker 1: The next use case is about supporting. So for all of us, there are a lot of situations in our life when we just need to talk to another human, but unfortunately it can be difficult. Perhaps it's such a topic on which you are not ready to discuss with, even with your closest friends.
Maybe this conversation actually needs to be, um, with a professional therapist or couch and realistically, not all of us can afford this sort of support. Speaker 1: So this is an opportunity for chatbots to really step up and this is just an example of another bot that we build that we call Cindy, whose goal is to help women and priority entrepreneurs and, um, and professionals to deal with this feeling of imposter syndrome.
Speaker 1: This minimization that you do of your own successes, achievements and talents and constantly feeling as if you're a fraud and they're about to find out that you're actually not good as you as good as you are. Speaker 1: So the solution here was, you know, ideally you would want to talk about it to your coaching therapist.
Speaker 1: So we can't quite give you that, right? Speaker 1: It's not scalable to the billion of users out there, but what we can do is that we can use this interface that you already personify and make it reflect and mirror to you, your own thoughts. Speaker 1: So it helps you basically to examine and refrain those thoughts that are holding you back.
Speaker 1: And this is just one example, again, of what chatbots can do so they can support you in all kinds of conditions that you're living with, or maybe, um, dealing with abusive relationships. Um, and there is really a variety of things that, um, you can use chatbots where you need this human touch. Speaker 1: And then finally, let's talk about service.
Speaker 1: And this is also a very important part where we don't necessarily need a human and maybe we don't necessarily need to be super engagement. Speaker 1: We just need information. We need information. We need clear instructions. And one good example is one of the bots that we've, uh, launched very, very recently. And the idea is how can small businesses can, uh, cope with this new regulation that relates to COVID-19.
Speaker 1: There are a lot of regulations now that businesses like restaurants need to comply with. And it's a, it has every regulatory system. It's pretty difficult to navigate. So what chatbot allow us to do is to really, um, provide the step-by-step instructions and situation relevant information so that people will, um, so that people will receive the information that's relevant to them.
Speaker 1: So this is just, uh, another bunch of examples of how, um, how chatbots can help you from applying to food stamps, to migrating for work, to reporting your experiences, uh, from sexual assault to bribery. Speaker 1: So I hope that by this point, I, uh, inspired you a little bit and I made you realize what a powerful tool we're having at the tip of our fingers.
And hopefully by, by now you want to, you're asking yourself, how am I going to go ahead and build my own chatbot. Speaker 1: So this is just a gist of how chatbot project is done. You start by really understanding your users, researching their needs and understanding what problem you're solving. You go, you go ahead and design the solution using a variety of conversational design tools.
You then iteratively in an agile manner, uh, implement your solution. Speaker 1: And one of the crucial points here is really the iteration part because chatbots are inherently give you user much more freedom than all the graphical solutions and graphical interfaces ever done. Speaker 1: They allow your users to, um, to put anything in the textbook from a curse, uh, to a thank you, to any, to any query that maybe you're not able to address.
So the importance of getting your chatbot in front of your users and repeatedly improving their experience by seeing how are they actually trying to interact with your solution is more crucial than with any other digital tool. Speaker 1: And finally, what we usually do with the organizations that we work with is that we pilot the solution for, for some period of time.
We try to understand whether the solution actually worked and was it a good solution to the problem that we were facing. And then the process of scaling is also, um, iterative, iterative, um, and keeping, keeping your head on the post all the time. So finally, here are a few chatbot do's and don'ts. I know that we're coming to the end of their time.
So I just want, I will let you stare at this just for a second. There are a lot of things that we see that we saw already just understanding your users, investing in your chatbot content, iterating and improving. And also one of the important things that you shouldn't do is try to do too much. So don't think that you can build a Siri, um, in, in a few days, understand what is it that your child should be able to do and do it really, really well.
And finally, don't leave it alone. See it as your childhood as a living being, it constantly needs a bit of upbringing and a bit of corrections. As it interacts and encounters new and new situations that it needs to cope with. So I hope that you took from this session today that first of all, chatbots are not just for companies and brands.
That they can engage, support and serve the most vulnerable corporations among us. And then finally here, it's not necessarily easy to build, it's easy to build a chatbot, but it's not easy to build a good one. So there are a couple of practices that you need to implement in order for chatbots to bring real impact. Thank you so much for listening and enjoy the conference.
Transcribed automatically; names checked against the speaker record. Select a time to jump there.
Article based on this talk
Why chatbots belong in the social impact sector, not just in customer service
Maria Dyshel shows how conversational AI helps nonprofits engage, support and serve the people who need them most, and why building a good bot takes discipline.
In short
Chatbots are not just for companies and brands. Maria Dyshel, CEO of Tangible AI, shows how nonprofits and social businesses use conversational AI to engage, support and serve vulnerable people at scale, from anti-trafficking education in Nepal to imposter syndrome support. Her warning: building a chatbot is easy, but building a good one takes user research, constant iteration and ongoing care.
Key takeaways
How nonprofits and social businesses use chatbots to scale their impact.
Real-world use cases of conversational solutions for good.
Do's and don'ts of creating a pro-social chatbot.
A double life that became a company
Dyshel's path into this work started like many tech careers: a computer science degree at the Hebrew University in Jerusalem, followed by algorithm and mobile development at Google and AI project management at Sanofi. Alongside that, she ran environmental campaigns, built technology for underserved populations with Engineers Without Borders, co-founded a nonprofit and travelled to India to help a local organization support rural nurses with a mobile application.
For more than a decade she lived what she calls a double life, and one question kept nagging her: "Isn't there a way to take something that I'm good at and use it to improve the lives of those that were less lucky than I am?" (1:50)
That question led her to found Tangible AI about two years before the talk, with a mission to put artificial intelligence into the hands of those working on humanity's most pressing problems. Working with her co-founder Hobson Lane and her team, she realised the answer had to be more holistic. Today the company consults on AI strategy and custom solutions for nonprofits and social businesses, builds its own products in an innovation hub, and runs an internship program to train a new generation of AI professionals who understand what it means to build prosocial technology.
Why conversation is such a powerful interface
Dyshel defines conversational AI simply: technology that lets people and computers communicate the way humans communicate with each other, through natural back-and-forth exchange, either by text with a chatbot or by voice with assistants such as Alexa, Google Home or Siri.
The first chatbot dates back to the 1960s, she notes, but until recently bots were mostly a gimmick. Several forces changed that. As she cited, messaging platforms now reach about 5 billion users every month, and in some rural areas people without electricity or running water still have phones they can message on. Younger generations prefer texting to calling, and digital life has made everyone more impatient for instant answers.
Underneath all of that sits human nature. However much the way we live has changed, we still exchange ideas through dialogue. "We still bond and organize using human language and using this interactive turn-based thing that we call human conversation. And this is why it's so powerful." (6:48)
She adds a psychological twist: pareidolia, our tendency to see faces and human shapes everywhere. "So when we encounter something like a chatbot, something that, as you say, walks like a human and talks like a human, our brain thinks that must be human." (8:04) That personification is a challenge for conversation designers, she says, but also an opportunity to scale the most powerful thing humans have and give billions of people better information and support.
Engage, support, serve: three jobs chatbots do for good
Dyshel organizes social impact chatbots around three functions, each illustrated with a bot her team built or launched.
Engage. As the speaker cited, about 35,000 Nepali girls and women between five and 18 are trafficked from Nepal to India every year. For teenagers in Nepali villages, risky situations are simply part of life, and the advice that might protect them rarely lands. "Cause let's admit it safety is boring. No one likes to talk about safety." (11:17) With Plan International, Tangible AI built Maya, a chatbot that turns safety and safe labour migration into interactive content with quizzes and choose-your-own-adventure stories, making it not only accessible but memorable.
Support. Some conversations people are not ready to have even with close friends, and not everyone can afford a coach or therapist. Cindy, another of the team's bots, helps women entrepreneurs and professionals work through imposter syndrome. It cannot replace a therapist, but it can use an interface people already personify to mirror their thoughts back, helping them examine and reframe the beliefs holding them back.
Serve. Sometimes people do not need a human touch or high engagement, just clear information. A recently launched bot helps small businesses such as restaurants navigate new COVID-19 regulations with step-by-step, situation-relevant instructions. Similar bots, she notes, help people apply for food stamps, migrate for work or report experiences ranging from sexual assault to bribery.
Building a chatbot that actually works
For viewers wondering how to build their own, Dyshel outlines the process: understand your users and the problem you are solving, design the solution with conversational design tools, implement it iteratively in an agile way, pilot it, and then scale carefully.
Iteration matters more for chatbots than for any graphical interface, because users can type anything, from a curse to a thank you to a question the bot cannot answer. "So the importance of getting your chatbot in front of your users and repeatedly improving their experience by seeing how are they actually trying to interact with your solution is more crucial than with any other digital tool." (17:02)
Her do's and don'ts follow naturally. Understand your users, invest in content, and keep iterating. Do not try to do too much: you will not build a Siri in a few days, so decide what your bot should do and do it really well. And do not leave it alone after launch; treat it like a living being that needs ongoing upbringing and correction as it meets new situations. Her closing line sums up the discipline required: "And then finally here, it's not necessarily easy to build, it's easy to build a chatbot, but it's not easy to build a good one." (18:46)
Watch the talk
Making the World Better, One Artificial Conversation at a Time: How Chatbots Are Powering the Social Impact Sector, by Maria Dyshel (19 minutes): https://www.womentech.net/video/maria-dyshel-making-world-better-one-artificial-conversation-time-how-chatbots-are-powering
About the speaker
Maria Dyshel is the founder and CEO of Tangible AI, a company building AI-driven and chatbot solutions for social impact organizations worldwide, and a Forbes 30 Under 30 social entrepreneur. She has held AI research, engineering and management roles across industries and has supported and led the creation of more than 20 chatbot applications across different sectors.
Questions this talk answers
What is conversational AI?
Why have chatbots spread so quickly in recent years?
How do nonprofits and social businesses actually use chatbots?
Why is iteration more important for chatbots than for other digital tools?
What are the biggest mistakes when building a pro-social chatbot?
Drafted with AI from the talk transcript. Editor review pending.
Frameworks and ideas mentioned
Engage, support, serve (chatbot functions for social impact)
Agile iterative development
Cite this talk
Maria Dyshel. "Making the World Better, One Artificial Conversation at a Time: How Chatbots Are Powering the Social Impact Sector." WomenTech Network, 2021. https://www.womentech.net/video/maria-dyshel-making-world-better-one-artificial-conversation-time-how-chatbots-are-powering