A short presentation I gave today at the Immersion 2011 event and building on the Delphi exercise at the Open University's ReLive 2011 to look at the future challenges of virtual worlds. Please read the notes to each slide for the detail.

One of the things we've been looking at is creating a motivation engine for our in-world bots based around Maslow's Hierarchy of Needs. Seeing as someone posted about this recently in Robitron we thought it might be an idea to share an early version of that we've done. Just click on "time passes" and the bots life moves forward by one time unit (15 mins). The bot then checks each of its drives (physiology, safety, social, esteem) and then chooses a task to address the highest need. Each need is given a weighting, so you can over-turn Maslow's hierarchy if you want, and for multi-period activities you can set whether the bot flits form task to task, or sees a task through to the end, or only changes if something else becomes critical. Each need is rated 0 - 100, and then each need is weighted 1-10. Safety is not tracked in this model as that is primarily driven by environment cues. Look at the source for a better idea of what is going on. All in all pretty Sims like, but gives us the basis of something to put into the bot. Warning: There may well still be bugs in it, is was developed as a test bed to try ideas out on (on my iPad).

---

---

We've finally decided to get a public sim on OSGrid in order to better showcase some of the work we are doing on OpenSim. We have our own internal OpenSim network (nicknamed DadenWorld), as well as individual developer worlds, but to make things more publicly available we decided it was simplest to go with a hosted service from Dreamland Metaverse on OSGrid.

The sim is called Daden Open - you should find it in your map search box on OS Grid. Our intention is not to build a specific island there, but rather to cycle through some of favourite OAR files - perhaps changing the island once a month or so.

To kick things off we've uploaded the refugee camp OAR file we did a few months ago for a trip to the UN. The sim is pretty passive but we've put in a simple information bot, and also showed the sort of question mark task markers we use with PIVOTE in SL and OpenSim to show how tasks might be identified and triggered.

Feel free to head over and take a look, and we'll let you know via the blog when we update the island with a new OAR file - probably you 20 parsec stellar neighbourhood model will be next up.

Independent analyst firm CCM predicts that "virtual agent implementations will be an industry standard in 3 to 5 years, a critical component in customer services and even the main point of contact in 2014".

The study, conducted in Q1 2011, surveyed 57 global leaders on their plans for using virtual agents as part of their customer service strategy. The group predicted:

• 100% increase in virtual agents in 2011
• 400% increase in virtual agents by 2014

Virtual Agents (or chatbots) are computer applications that can act as a first line of support providing instant responses to questions, via your website and facebook pages, without the need for staff intervention. They can answer queries, using natural language, about your organisation and services, reducing pressure on your resources.

A majority of the respondents were interested in using the virtual agents for "right channeling" purposes as well as:

• integration of the virtual agents with live chat (66%)
• as first point of contact for all online customer interactions (56%)

This latest report continues to adds independent documentary evidence of how using a natural language chatbot can be cheaper and more effective than some of the alternatives. Indeed the survey also captured reasons for using virtual agents /chatbots which included:

• 67% note eDeflection from higher cost customer service channels
• 51% believe that virtual agents will increase their "customer requests" handling capacity.
• 36% believe virtual agents will provide insight into user trends to determine and optimize new service issues as they arise.
• 33% believe that VAs will address search limitations.

Daden have always advocated the advantages of using chatbots, re-releasing their own white paper on the subject last year. This latest report adds to the increasing number of independent studies and continues to reaffirm that virtual agents, or chatbots, are here to stay!

A chatbot or virtual receptionist is a computer-controlled application that has the ability of answering questions about your company, products and services immediately. 

Here is the result of a recent conversation that our chatbot Abi, had with someone visiting our website:

User: what are you

Abi: I am chatbot or virtual customer assistant designed to help visitors to find out more about Daden

User: what is daden
Abi: Daden is a virtual worlds and virtual characters solution provider.

User: can you learn
Abi: I can only learn by being programmed at the moment

User: what can you do
Abi: I work at Daden's offices in central Birmingham and also on our island Daden prime in second life

User: what is second life
Abi: Second life is a virtual world, one that only exists on the servers of linden labs, and in the browsers and minds of its users

User: can i get a chatbot
Abi: If you would like to get a quote for a chatbot for your web site or virtual world presence please email sales@daden.co.uk and we will discuss your requirements with you

User: thanks
Abi: Good

User: byb
Abi: Sorry, I only have limited knowledge. I can only talk about Daden

User: bye
Abi: Bye. I'd be grateful if you would fill in a short survey on me. Link below.

Daden check Abi's conversation logs frequently and update her regularly and this conversation is a good example that shows how people do build a rapport with a chatbot and are willing to talk to it - and thank it too!

Daden have given Abi a personality outside of her main duties - which are to talk about Daden and to point people to pages on the website. Here's an example:

User: i want us to be best friends.
Abi: My best friend is Halo Rossini but I do have other friends

User: can i be your friend too?
Abi: Oh dear. I'm sure we could be but I'm at work I'd rather concentrate on answering questions about Daden's services and products

User: so you keep saying, but i want to chat about something else.
Abi: I like to talk about what we do at Daden

User: yes i know, but i want to talk about how lovely you are.
Abi: Say again? I only have limited knowledge. I can only talk about Daden 

Most organisations we work with choose not to give their chatbots a personality but use their chatbots to act as a virtual assistant available 24hrs a day to guide people around their website or act as an FAQ. Research in the US (where chatbot use is more advanced) has shown that the average cost-per-contact using a chatbot is under 60p compared to over £3.50 for live text-chat or over £4 for a telephone contact. Leading analyst Forrester reported in 2010 that chatbots "offer compelling business benefits, including enhancing customer experience, reducing live help costs, and driving cross-sells".

Chatbots can be placed on websites, on Facebook and inside virtual worlds. For example, Abi also lives on Daden's island, Daden Prime in Second Life, where she acts as a virtual receptionist and answers questions. One of our clients has their chatbot answering questions by SMS, email as well as via their website. It's multi-lingual too answering questions in both English and Arabic!

Talk to Abi, our Virtual Receptionist yourself here.

Daden have a white paper available to download titled: Deploying Chatbots to Customer Advantage

Click on this link to download your copy

If you'd like to learn more about chatbots or have visitors thank your website email soulla.stylianou@daden.co.uk  or call us on 0121 250 5678.

We've been making good progress over the last month, and our toolset around data plotting in immersive spaces is taking shape. What has emerged are four main plot types, and supporting tools, all just importing data from CSV files to a standard format:

The simplest is a surface plot, aka a 3D bar chart. Data consists typically of one dimension of entities (e.g. stock ticker symbols), one of time (e.g. years or months) and one of value (e.g. stock price). Controls are provided to be able to change transparency levels, zoom in and out, and switch labels on and off.

Next is the scatter plot. In this tool we actually implement a 9D data cube, and allow you to plot any dimension not only against the 3 axes (x,y,z), but also to colour, size and shape. The tool supports the ability to zoom the data in and out - as well of course as your ability to fly in and through the data.

Using the same tools as the scatter plot we can also implement social graph mapping. here we can show relationships between people, places and organisations, and colour and size code links dependent on their detail - as well as applying similar marking to the data points.

The final version is again a special case of the scatter plot (what we term generically a static plot), where x and y are lat and long (or other geo co-ordinates) and we plot the data against a map - in this case location data derived from Twitter. Here we also colour code the points by recency, and shape them by static or mobile browser. The white layer is people mentioned by geocoded tweets, and the yellow boxes the hashtags.

Finally to bring it all together we captured a load of tweets during the recent Royal Wedding and this is what we got:

Based on over 5000 tweets captured at 5 minute intervals over the day. Central cone layer shows most prolific tweeters (size = number of tweets, red = quoting others, yellow=being quoted). Position in layer reflects sentiment, high=good, low=bad (orange surface level = neutral). Lower cone layer is other Twitter users being mentioned in tweets, upper cube layer is hashtags being used, for both size=number of mentions. Each object has a label identifying name. Hovering cursor over an object brings up more information/options such as bringing up the twitter page of the user (in this case imPrinceHarry - who luckily rated positive on sentiment!). Further refinement could show social graph data from people mentioning other people in tweets, and links to
hashtags they used etc.

Our little adventure is over yet though. With the release of OpenSim 0.7.1 we've finally been able to start writing Region Modules (and delete times are back to reasonable), getting a four fold increase in plot speed and more robust comms, so we're now moving out plot engine to RegionModules (although we'll keep a LibOMV option for remote deployments). We've also begun to look at what it will take to port the whole thing to Vastpark and Unity.

If you'd like a demo, or to learn more about our work in data visualisation or visual analytics, or have a project which could use these techniques, then don't hesitate to get in touch.

One thing we've found hard over the last few months is to get people in the architecture/building sector to understand how what we do with the built environment (e.g. with the Birmingham Library project) is different to what they do. For architects in particular it appears to be all about "look" - the better a computer image is to perfection (not reality) the better it is - regardless of the use. In our recent Built Environment white paper we presented the "whole life" model of a building that can be supported by virtual worlds - where the "social" dimension of the build is paramount.

This has led us to think that maybe the right message to get across is that there are 3 "models" that anyone involved in the built environment should be concerned with:

• the architects visual appearance model - what will it look like
• the engineer/builders structural model - how do we build it
• the social model - how do we use it

At present there is a lot of debate about Building Information Modelling (BIM) - the overlap between architects and engineers, but the Building Social Model (that which virtual worlds are ideally suited for) is getting overlooked. In the future (10Yrs+?) hopefully one platform will cover all three uses, but at the moment we need to find ways in which projects can explore all three dimensions, probably across 2-3 tools, but in the most cost-effective manner.

Looks like the IGI book " Multi-User Virtual Environments for the Classroom: Practical Approaches to Teaching in Virtual Worlds" that we wrote a PIVOTE chapter for is out. The chapter explores the background to authoring learning systems in virtual worlds, why we developed PIVOTE, the PIVOTE architecture and an example use case, and a couple of case studies from institutions using PIVOTE. Rest of the book should be good reading too. Enjoy!

Get your library to buy it here - http://www.igi-global.com/bookstore/titledetails.aspx?titleid=47435

Buy just the PIVOTE chapter as a PDF here - http://www.igi-global.com/bookstore/chapter.aspx?titleid=53498