Recently in StudentProjects Category

Brief overview/background of the project:

One of the key differences between virtual world training/education scenarios and "traditional" eLearning scenarios is that the learner is represented on-screen by an avatar. How much does this on-screen presence contribute to learning and retention, and how is that balanced by the extra time taken to learn how to control the avatar.

Aim of the project:

Conduct A/B comparisons between the same learning exercise conducted in an (A) environment where the user has no avatar and a (B) environment where it does have an avatar - in order to see how much the avatar adds or detracts from the learning.

Possible Approach

One approach would be to make/source an existing no-avatar 2D style (but graphical) learning exercise and recreate it closely in a 3D avatar based platform and then compare the two - both in understanding of content, and recall of content some time later. Another approach would be to use a flexible 3D avatar platform (eg Second Life) to create both the avatar and non-avatar scenarios - so that the only difference is the avatar presence/non-presence and the simplification of control that results.

A clear example of something similar or in the field:

There are a large number of 3D avatar based environments (eg Second Life), and even some 3D/no avatar ones (eg some Web.Alive), and a lot of 2D environments without avatars.

Skills Needed:

Whilst some technical skills would be required (eg some scripting and 3D build) this is primarily a usability/HCI assessment with aspects of learning psychology.

What the student might get out of the project:

Time plan:

The project will probably take 1 -2 months to plan and do a literature search, 1 - 2 months data collection (maybe 3 months if doing experiments), and 1 -2 months to write up.

Overview: It is very easy for a computer to do mathematical calculations. However when we try to create a bot which reflects human capabilities we want the bot to show only a human level of mathematical ability - so it could tell you say 10 / 4 but not 1924/23. What we would like to incorporate into our bots is a set of algorithms (ideally in C# or Perl, but pseudo-code fine) which delivers a human level of mathematical ability.

Educational Aim: The aim is to create a set of algorithms (ideally in C# or Perl, but pseudo-code fine) which would enable a bot to emulate a human level of mathematical capability.

Of course an initial part (and probably the key part) of the project would be to decide what a "human-level" of mathematical ability actually is - and how it can be characterised.

There is also the additional issue of how the bot recognises and categorises a mathematical request (eg "what is six times nine"), and how the result is given back in a naturally human way (eg "45 I think")

Note: This same need to "dumb-down" a computer to emulate a human of course applies to many others of computer ability which could form other projects (eg mis-spelling, location identification, periods of time etc)

Example: Unknown

Skills: This is primarily a software task but there are also some interesting issues of educational and straight forward psychology about how people learn and do maths and relate to numbers.

Student Benefits:

1.Working with and to a brief set by a leading-edge commercial company
2.Looking at an instance of a generic issue in the future of AI and virtual assistant development, and an understanding that creating an "imperfect" computer is harder than creating a "perfect" computer
3.An insight into the way that the human brain processes tasks in a radically different way to a computer
4.Seeing their algorithm implemented in avatar/virtual world and text-to-speech technology

Time Plan (elapsed):

1 - 2 months research to establish "human maths" levels
1 - 2 months to extract the characteristics of human maths
1 - 2 months to code
1 - 2 months to test with real users (a mathematical Turing Test?)

Brief overview/background of the project:

As with any experience the first 5 minutes can often be crucial to a users eventual use of or satisfaction with a system. The first 5 minutes in a virtual world is notoriously disorientating and problematical. If we can make this first 5 minutes a better experience, by better tuning it to the conscious and unconscious needs and emotions of the user then we can increase not only their satisfaction and productivity but also the broader uptake of virtual worlds.

Aim of the project:

To identify what a user experiencing during this initial period in sensory motivational and emotional terms, and how the experience can be made more productive.

A clear example of something similar or in the field:

This is a classic user-centered design issue and there is much literature on 2D systems and should be some on 3D systems. Game design will also be relevant - as might real-world orientation issues.

There are several "orientation" areas in Second Life which could be studied, and experimental ones created. Other worlds have other approaches and some dispense with orientation altogether.

Skills Needed:

There are no specific technical or domain skills required. What is important is a good analytical process and mind, with a good data collection strategy, and insightful analysis. A usability, design or psychology background might be useful.

What the student might get out of the project:

A good understanding of user design issues within 3D spaces.

Time plan:

The project will probably take 1 -2 months to plan and do a literature search, 1 - 3 months data collection and experiments, and 1 -2 months to write up.

Brief overview/background of the project:

Virtual worlds give us the opportunity to not only plot data in 3D but also to move our avatar inside it. With spatial data being plotted on a map we can literally move our avatar to the data point or location being talked about. With 3D visualisations of DNA we can have a meeting sat on a particular gene or base. There is a lot of anecdotal evidence that this ability to get inside data and relate to it spatially as we do in the real world is a real benefit of virtual worlds. Somehow we do not "see" the data as a 2D projection on the VDU screen, but rather we "see" it through the eyes of our avatar. But most plotting of data either reflects 2D metaphors (pins on maps, lines on graphs), or is a literal 3D model (aircraft on approach to LAX). However in a virtual world the data could be represented as anything. A FTSE100 time-series could become a forest growing and dying over time, the spread of a disease could be eruptions of mushrooms or the stalking of robots.

Aim of the project:

To find out whether certain non-standard data representations aid or hinder understanding and recall of data sets. Are different styles better suited to different sorts of data or analysis. Do the results vary with size or complexity of the data set - or the personality and experiences of the user?

A clear example of something similar or in the field:

2D metaphor has a long history in data visualisation - Edward Tufte is probably a key reference here.

Skills Needed:

There are no specific technical or domain skills required. What is important is a good analytical process and mind, with a good data collection strategy, and insightful analysis. A statistics, social science or psychology background might be useful.

What the student might get out of the project:

The student should gain a good understanding of 3D data representation issues, which may also be applicable to 2D representation.

Time plan:

The project will probably take 1 -2 months to plan and do a literature search, 1 - 3 months data collection and experiments, and 1 -2 months to write up.

Brief overview/background of the project:

Virtual worlds give us the opportunity to not only plot data in 3D but also to move our avatar inside it. With spatial data being plotted on a map we can literally move our avatar to the data point or location being talked about. With 3D visualisations of DNA we can have a meeting sat on a particular gene or base. There is a lot of anecdotal evidence that this ability to get inside data and relate to it spatially as we do in the real world is a real benefit of virtual worlds. Somehow we do not "see" the data as a 2D projection on the VDU screen, but rather we "see" it through the eyes of our avatar. Is this true? And is so does it bring benefits to understanding or to recollection.

Aim of the project:

To examine how a user relates to data within a 3D space, and in particular whether they have a spatial relationship to that data based on the relative location of their avatar - and whether such a relationship, if it exists, helps with the understanding and subsequent recall of the data and related discussions.

A clear example of something similar or in the field:

3D data visualisation without avatars has been around for some time and should be covered in the literature, but there may not be much on immersive data environments.

Skills Needed:

There are no specific technical or domain skills required. What is important is a good analytical process and mind, with a good data collection strategy, and insightful analysis. A psychology background might be useful.

What the student might get out of the project:

We expect 3D data visualisation to be a key market in the next 5 - 10 years and this will give the student both an insight into the issues involved, and a better understanding of psycho-visual perception.

Time plan:

The project will probably take 1 -2 months to plan and do a literature search, 1 - 3 months data collection and experiments, and 1 -2 months to write up.

Brief overview/background of the project:

This is a literature search project. 3D computer simulation has been around for over 20 years. Initially simulation was costly and the domain of defence and aviation industries. But with the rise of game based technologies the scope for cost-effective training simulation has increased. But what are the quantifiable benefits of training in simulations over training in real-life. And is an element of real-life still needed?

Aim of the project:

To conduct an academic and commercial literature search on the quantifiable benefits of 3D immersive training, focussing particularly on serious games and virtual worlds, but also looking at the broader bespoke 3D training system marketplace.

A clear example of something similar or in the field:

n/a

Skills Needed:

There are no specific technical or domain skills required. What is important is a good analytical process and mind, with a good data collection strategy, and insightful analysis.

What the student might get out of the project:

A good understanding of the virtual training market and the commercial rationale for such systems.

Time plan:

The project will probably take 1 -2 months to plan and do a literature search, 1 - 2 months data collection, and 1 -2 months to write up.

Brief overview/background of the project:

Traditionally users have known when they are talking to a chatbot on the web. However in a virtual world there need be nothing to signify whether an avatar is being driven by a bot or a person. Whilst the quality of the conversation is then undoubtably key to deciding whether an avatar is a bot or a human what role does the appearance of an avatar take. If a bot looks like a bot will users trust it more? If a bot looks like a sexy male or female will users pay less attention to the unhumanness of the replies - or will they regard it as deceitful? And if a human takes on a bot avatar are they seen as more bot like?

Aim of the project:

To analyse how users draw conclusions about whether an avatar is a human or a bot from the appearance of the avatar - and whether conflicts between nature and appearance cause problems for the user.

A clear example of something similar or in the field:

There should be a certain amount of literature on this from stand-alone projects and possibly from MMORPGs.

There are probably two key approaches to this project, both of which involve staging a mini Turing Test:

Use a real chatbot, each bot talking to the same bot, but varying the appearance (bot and non-bot)
Use only real people ( a Wizard of Oz study) but varying appearances (bot and non-bot)

This project could also be linked with the one about avatar appearance choices and assumptions for human users.

Skills Needed:

There are no specific technical or domain skills required. What is important is a good analytical process and mind, with a good data collection strategy, and insightful analysis. A social science, psychology or even fashion background might be useful.

What the student might get out of the project:

A good understanding of the importance of visual appearance to perception - which may well have implications and cast new perspectives on real world issues such as prejudice and equality, as well as in game and interface design.

Time plan:

The project will probably take 1 -2 months to plan and do a literature search, 1 - 2 months data collection (maybe 3 months if doing experiments), and 1 -2 months to write up.

Brief overview/background of the project:

The UK Citizenship tests provides a useful collection of the "general knowledge" of a UK citizen. It also opens the (slightly tongue-in-cheek) question of "if a bot could pass the test then should the bot become a citizen"

Aim of the project:

To create chatbot which can be asked any question based on the knowledge of the UK Citizenship and correctly answer enough questions to achieve a pass mark.

A clear example of something similar or in the field:

There are lots of chatbots designed to provide "virtual assistant" services - eg Anna on the Ikea web site.

There are two possible approaches - one "quick and dirty" the other more rigorous but both would be valuable learning exeperience.

- Use the AIML style input/response method of building a chatbot
- Use an RDF like approach to encode the answers as "knowledge" and then query them

Daden's Discourse chatbot engine supports both methods.

Skills Needed:

There are no specific technical or domain skills required. What is important is a good analytical process and mind, and the ability to appropriately analyse the data implicit in the questions and answers. Training in AIML/RDF can be given. All authoring is on web forms.

What the student might get out of the project:

A good understanding of knowledge representation, the creation and testing of chatbots, and potentially ethical issues around citizenship!

Time plan:

The project will probably take 1 month to plan, 1 - 2 months data capture and authoring, 1 - 3 months testing and and 1 -2 months to write up.

Brief overview/background of the project:

In trying to develop AI systems within virtual worlds we are faced with the issue of what information should a bot remember. The bot is potentially sensing many kilobytes of data a minute (object movement, avatar movement, chat conversations, weather etc). We could just store all of this, but in developing an AI a possible avenue is to store (and retrieve) memories in a human-like way. Simple personal experience tells us that we don't appear to remember every change in our environment that we see/hear/feel, so how does the human brain handle this?

Aim of the project:

To examine how humans filter current (and especially peripheral) experience into memory.

One major issue is that of attendance. As soon as I think "will I remember this" the chance are that I will remember it better. If I sit in a busy coffee shop and and am then asked about what was happening whilst there, 1 minute/hour/day/week later how much will I remember. Does a sleep cycle introduce a major fall off? How unusual must an action be to be remembered separately from the noise? To avoid the attendance effect we can also probably only ask each subject once!

A clear example of something similar or in the field:

Work on peripheral awareness and subliminal advertising may be of interest. There is also a large body of work on memory and different memory models.

Skills Needed:

There are no specific technical or domain skills required. What is important is a good analytical process and mind, with a good data collection strategy, and insightful analysis. A psychology/neuro-science background might be useful.

What the student might get out of the project:

We think this is a key issue in AI and human memory which for an innovative approach so could have good paper potential - and even become an accepted standard. The understanding could be applicable to a wide variety of psychology issues and disorders, as well as in AI development. There may also be ethical and other issues of experiment design which would be of particular benefit.

Time plan:

The project will probably take 1 -2 months to plan and do a literature search, 1 - 3 months data collection and 1 -2 months to write up.

Brief overview/background of the project:

A quick analysis of the 2009 Loebner Prize (Turing Test) transcripts shows two key features:

All the human confederates made and/or corrected spelling mistakes, whilst each computer typed perfectly
All the human confederates regularly made multi-line responses, whilst none of the computers did

So quite aside from any content, we could 100% predict which was the computer and which the human by such an analysis.

Aim of the project:

The aim of the project is to find out how important such peripheral information as spelling mistakes and multi-line responses are in convincing a judge of the human-ness of an entity. Is it something they consciously notice, does it subconsciously inform their decision or is it ignored. Such information would be valuable in deciding the best strategy to pass the Turing Test and to create a "human" chatbot.

A clear example of something similar or in the field:

There has been much discussion on the Robitron list about the ethics and importance of such "tricks" in chatbot design. There may well also be papers in the literature.

Skills Needed:

There are no specific technical or domain skills required. What is important is a good analytical process and mind, with a good data collection strategy, and insightful analysis. A psychology background might be useful.

There are probably two key approaches to this project, both of which involve staging a mini Turing Test:

Use a real chatbot, each bot talking to the same bot, but varying the bots use of peripheral signals
Use only real people ( a wizard of Oz study) and get them to consciously vary the peripheral signals they offer

What the student might get out of the project:

A good understanding of the importance of peripheral signals to our perception - which may well have implications and cast new perspectives on real world issues of communications.

Time plan:

The project will probably take 1 -2 months to plan and do a literature search, 1 - 2 months data collection (maybe 3 months if doing experiments), and 1 -2 months to write up.