How Hot are Chatbots?

bots, , , ,

The last session that I attended at the Legal Services Corporation’s Innovations in Technology Conference last week was entitled ‘Chatbots are so hot’.

It covered presentations on ATJ  Bot on Tennessee’s Help4TN website, coloured blue and also Chicago’s Rentervention bot, coloured red as above and previously covered in an earlier post. For a review of global progress on chatbots, you could throw in the experience of Scottish Shelter with Sheldon and Alisa (both now defunct), Joshua Browder’s DoNotPay suite (apparently blooming) and one of the winners of the first round of the NESTA Legal Access Challenge in England and Wales, Mencap’s Aeren. Also around, but not considered here, are the more commercially orientated LawDroid and Robot Lawyer Lisa.

Let us begin with a definition. A chatbot is, and I am quoting Lawdroid’s site, ’a computer program designed to simulate conversation with human users, especially over the Internet’. This started to get traction with the launch of Cleverbot in 2011 which was the first computer programme that passed the ‘Turing’ test albeit by a small majority – just over half of those making a judgement on it were fooled into thinking it was human.  Chatbots are now pretty widespread in the commercial market with Siri and Alexa as two well known examples of speech based bots. Banks and other consumer businesses make great use of bots in conversing with customers. In 2016, Facebook launched a platform to make chatbots available to those on Facebook Messenger and that has spawned thousands. Probably the most famous user in the access to justice field was then Stanford student Joshua Browder who launched his DoNotPay bot to challenge parking tickets. 

Bots can vary greatly. Most are little more than guided pathways that take a user through a series of questions and answers. Some claim to use natural language processing and/or machine learning to identify questions and to improve answers. Most chatbots are little more than chat boxes – blobs that contain an entry from the user or the bot. Most have no visual identity though Nadia, which had a lifelike face that spoke with the voice of Cate Blanchett, was a brief glorious exception until she fell foul of government funding cuts and the shortcomings of the then available artificial intelligence generated by IBM Watson. She was created by New Zealand company, Soul Machines. For a rather chilling sense of what the commercial future holds, consult their website and note: ‘Soul Machines™ is the only technology solutions provider that offers complete capabilities for brands and companies revolutionizing human experience. Now you can create Digital Heroes rapidly, scale infinitely, and drive customer and brand experiences like never before.’

Most chatbots are considerably less ambitious. Few attempt to do much to create a personality – though Mencap’s Aeren is an exception. It is her job to introduce you to the world of those with a learning disability. She likes dancing, drinking (though not, heaven forfend to excess) and Avril Lavigne. You do not have to go far in exploring conversation with her to get into sex as she takes you through a set of choreographed questions and answers about the experience of someone with learning difficulties.

Aeren’s role is largely promotional for Mencap’s concerns but Rentervention is in the business of giving direct information on housing problems, including eviction. You can see a short video on the programme here. And, if you make up a Chicago zip code then you can test the programme yourself. I used 60608. If you do this, you can compare Rentervention to the more conventional information script on the same topics that you can see by, for example, entering eviction as a question on English sites like citizensadvice.org.uk or the housing campaign organisation Shelter.

As an example of the linear approach, these are the headings that Shelter gives you for eviction:

Eviction notices

Private tenants and lodgers

Council and housing association tenants

Illegal eviction

Rent arrears

Prisoners

This is pretty uncompromisingly technical. For example, I have no real idea what a s21 eviction or a s8 notice is. It was all different when I last advised on housing. By comparison, Rentervention takes you through a set of short questions beginning with whether your landlord lives in the building and whether it has at least seven units. Get through this gateway and you get told: 

Great, you qualify to use Rentervention! First, I am a bot and not a lawyer. I can only give you legal information. I’m going to be asking you a lot of questions. This will help me do certain things for you (like help you create a letter for your landlord or give you tips for court). Sometimes, I will see if you qualify for a free lawyer and ask if I can give your information to them. The more you stick with me, the more I may be able to help you.

For this purpose, the content does not really matter. The point of the comparison is the difference the bot can make in digesting and transmitting information. Go back to help4tn and you can experience a bot which communicates verbally – way short of Nadia’s visuals (she could even see your expression through the computer’s camera) but offering an alternative to written communication.

So, what’s the answer? Are bots hot? Do they offer an improvement on linear written communication in the way of most information/advice sites? You make your own judgement on that. But, for myself, I would say that they have done enough to prove their potential. They are rather more complicated to run than a linear site and it is interesting that both Shelter Scotland and Nadia’s creators pulled out of their creation. But, particularly as commerce expands on their use, they must be fertile territory for more exploration in access to justice.

Leave a Reply