Ever since Alan Turing asked his famous question “Can machines think?” in the early 50s, people have never stopped considering the question from different angles. Philosophers tried to answer whether machines can have consciousness, neuroscientists explored their ability to acquire intelligence, computer scientists tried to mimic the human brain’s logic as far as they could, and filmmakers produced all kinds of movies about the subject, most of the time showing how machines can take over the world after outsmarting humans. However, one thing is certain, natural human-machine interactions have always been the ultimate test that machines must pass before we grant them the title of “rivals”, be it under the label of intelligence, consciousness, or any other fancy term.

Thanks to the outstanding success of Deep Learning in the last decade, we have witnessed a remarkable growth in the development of virtual agents capable of conducting increasingly complex conversations. These agents are commonly referred to as chatbots, virtual assistants, conversational systems, or even voice assistants when they have a voice component. Sometimes, they can even have a ‘body’, like Amazon Alexa or Google Home assistants. The bottom line is that we’re talking about interactive artificial agents that try to imitate human conversations.

The need for virtual assistants

If you think about it, chatbots are taking Human-machine interactions to the next level, the same level Humans use to interact with each other. Maybe just the last one before we discover how to implement telepathy 😎

Humans used to interact with machines through complex black and white coding interfaces, then we invented more human-readable programming languages to communicate our thoughts to machines. After that, we reached a stage of even more human-friendly interfaces based on buttons, web pages, and web applications. Over the years we have created simpler intermediary ways of communicating with machines, but now we’re knocking down this intermediate wall, we are just using our own native language(s), and it’s time for the machines to go the extra mile.

Being able to conduct natural, smooth, and efficient conversations has opened the door to countless applications for these chatbots. Particularly, they often offer a better alternative to deliver services such as personal assistance, news, customer service, entertainment, as well as wellbeing and emotional support.

As a matter of fact, this latter application is a perfect example of where chatbots can greatly improve peoples’ quality of life [1, 2]. For instance, the Mental Health Foundation is now considering stress-related disorders as one of the biggest burdens on healthcare systems worldwide [3]. Yet, a serious barrier towards providing psychotherapy to all those who need it is that there simply aren’t enough trained therapists to meet the escalating demand. Last year (2021), the number of psychologists in the UK was approximately 37300 [4], while the NHS claims that 1 in 4 adults and 1 in 10 children experience mental illness [5]. On the other hand, people face barriers that prevent them from starting therapy sessions: lack of time, cost, or embarrassment to name a few.

Here comes the power of virtual assistants. While they can’t replace human therapists, they can certainly be scaled, personalized, and be available anywhere, anytime. Studies [6, 1] have shown that they can be an efficient tool for extracting high-quality contextual and personal information that can help people detect, understand, and eventually improve their behaviors. They also present a useful companion for supervising patients in their coping and behavioral exercises [7, 8]. Not to mention that talking to a bot makes some conversations much easier [9]. People generally don’t feel stupid interacting with machines, asking the dumbest questions. If you disagree just take a look at your browser history 😏

Over the years, many movies and TV series have explored these AI-based virtual assistants, often to show the dark side of this technology and how things can quickly degenerate, leading eventually to undesirable outcomes. Black mirror’s “Be right back” episode is an impressive, yet fairly realistic, illustration of this point of view. Another exceptionally mature presentation of this technology is shown in the movie “Her”. It depicts a sensitive emotional story in a not-too-distant future of a man trying to rebuild himself after a heartbreaking personal experience, with the help of an AI-based virtual companion.

Happibot: Fiction vs Reality

Now the question is: Is this something we can create in our real world? Or is it just science fiction? Obviously, until today, we haven’t developed agents that are able to live up to the level of complexity portrayed in these movies, we’re not infinitely far beyond though, a couple of technical breakthroughs and we’ll be there 😀 And by “we” I mean Humans (no offence future AI readers 🙄 ).

To give you an idea, Happence, a leading company in the domain of mental health and wellbeing, has started to introduce their patients to a new product based on conversational AI: HappiBot. This AI assistant is already helping people unlock the power of a happy and healthy workforce, and improve their wellbeing. The system is tailored to each user, offering engaging conversations to better understand their moods as well as what led them to their different emotional states. It follows them throughout their social and emotional life, listens to their particular ups and downs, and suggests personalized resources that may improve their wellbeing.

Behind the scenes, HappiBot involves a task force of wide and diverse profiles. At AffinitiAI, we are taking part in this project by providing the expertise of our AI specialists, conversation designers, researchers, and software developers. Additionally, the team involves domain experts such as therapists and neuroscientists. This multi- disciplinary team is essential to provide both evidence-based and human-like interactions.

Besides what has already been said about intimacy and availability, one of the major benefits of products like HappiBot, is their ability to deliver human expertise in a scalable and affordable way.

For Happibot to achieve its goals while assuring engaging interactions, its conversations are made up of three different, yet coherent, layers of communication. First, the bot engages in a social chat aiming to create a friendly relationship (otherwise known as a therapeutic alliance) with the user. Then, it follows up with a Q&A exchange to extract the user’s current mood as well as the reasons behind them feeling that way. Finally, the recommendation engine processes all the collected information, together with the past data that the system knows about the user to come up with the best instructions to improve the user’s wellbeing.

Although the data collected by the chatbot is completely anonymized, analyzing it in its entirety opens unprecedented opportunities for experts and researchers. Studying this data leads to detecting useful patterns, knowing better what drives the moods of the patients, and advancing the related research fields.

Apart from the building blocks of logic governing the operation of the chatbot, the virtual assistant obviously needs a user interface to interact with people. For that, HappiBot provides a friendly web UI, as shown in the image below. This UI not only offers a way to communicate with the chatbot but also provides a complete platform containing all the resources users may need to improve their wellbeing.

We are not here debating whether machines will replace humans, whether they are conscious, caring, or sympathetic. We’re just talking about them “looking like so”. We’re talking about them being useful and able to help humans in their social and emotional quests. After all, scholars even argue about proving that other humans have consciousness, let alone machines 😅

Conclusion

As we’ve seen before, the idea of creating artificial personal companions has always fascinated humanity. Actually, ever since people started programming machines, they rapidly started creating rule-based chatbots. Agents that are able to answer specific questions using FAQ lists. However, if we’ve made the biggest leap towards flexible human-like assistants, it’s mainly thanks to the latest advances in Deep Learning (DL), especially Natural Language Processing, the branch of AI specialized in studying (wait for it …) Natural Languages. Unlike the first category, DL-based chatbots are able to learn from past experiences, understand the current context, and generalize beyond the seen data. No matter what users type, they are always able to assess the situation and ‘guess’ what should be said next.

Virtual assistants are no longer some futuristic idea depicted only in sci-fi movies. Millions of users are already using different types of general-purpose assistants like Alexa, Siri, Cortana, and Google Assistant. Nevertheless, in this post we explored the need for more targeted assistants, virtual experts so to say. HappiBot is a fitting example in the mental health & wellbeing domain

If you want to learn more about how Conversational AI can help you scale your business then get in touch with us here!

References

1. Kocielnik, Rafal, Daniel Avrahami, Jennifer Marlow, Di Lu, and Gary Hsieh. 2018. “Designing for Workplace Reflection: A Chat and Voice-Based Conversational Agent.” Pp. 881–94 in Proceedings of the 2018 Designing Interactive Systems Conference, DIS ’18. New York, NY, USA: Association for Computing Machinery.
2. Park, SoHyun, Jeewon Choi, Sungwoo Lee, Changhoon Oh, Changdai Kim, Soohyun La, Joonhwan Lee, and Bongwon Suh. 2019. “Designing a Chatbot for a Brief Motivational Interview on Stress Management: Qualitative Case Study.” Journal of Medical Internet Research 21(4):e12231. doi: 10.2196/12231.
3. MHF. (2016). “Fundamental Facts About Mental Health 2016”. Mental Health Foundation: London. Retrieved May 18, 2022 (https://www.mentalhealth.org.uk/statistics/mental-health- statistics-uk-and-worldwide).
4. Frédéric Michas. (2021). “Number of psychologists in the United Kingdom (UK) 2010–2021”, Statista. Retrieved May 20, 2022 (https://www.statista.com/statistics/318869/numbers-of- psychologists-in-the-uk/)
5. National Health Service. (2022). “Mental Health”, Retrieved May 20, 2022. (https:// www.england.nhs.uk/mental-health/).
6. Xiao, Ziang, Michelle X. Zhou, Q. Vera Liao, Gloria Mark, Changyan Chi, Wenxi Chen, and Huahai Yang. 2020. “Tell Me About Yourself: Using an AI-Powered Chatbot to Conduct Conversational Surveys with Open-Ended Questions.” ACM Transactions on Computer-Human Interaction 27(3):15:1–15:37. doi: 10.1145/3381804.
7. Fitzpatrick, Kathleen Kara, Alison Darcy, and Molly Vierhile. 2017. “Delivering Cognitive Behavior Therapy to Young Adults With Symptoms of Depression and Anxiety Using a Fully Automated Conversaional Agent (Woebot): A Randomized Controlled Trial.” JMIR Mental Health 4(2):e19. doi: 10.2196/mental.7785.
8. Inkster, Becky, Shubhankar Sarda, and Vinod Subramanian. 2018. “An Empathy- Driven, Conversa6onal Ar6ficial Intelligence Agent (Wysa) for Digital Mental Well- Being: Real-World Data Evalua6on Mixed-Methods Study.” JMIR MHealth and UHealth 6(11):e12106. doi: 10.2196/12106.
9. Lee, Yamashita, Huang, and Wai Fu. 2020. “I Hear You, I Feel You: Encouraging Deep Self- disclosure through a Chatbot”. Proceedings of the 2020 CHI Conference on Human Factors in Compu6ng Systems. Associa6on for Compu6ng Machinery, New York, NY, USA, 1–12. h)ps:// doi.org/10.1145/3313831.3376175