Using Double Agents as Electronic Tutors

new learn

Even the best-designed eLearning courses may benefit from individual coaching. But teaching staff costs money.

Video conferencing and pre-recorded lectures are one possibility. Animated 3-dimensional characters and adaptive programming can provide a cheaper solution.

The Double Agent Network

Microsoft Agent is a set of software services which present user-controlled applications as interactive personalities – animated 3D cartoon characters. Character-based interactions can be blended with conventional interface components like windows, menus and controls. Using text-to-speech and speech interfaces, spoken language input and sound effects can be added to the mix.

Double Agent is an Open Source alternative to Microsoft Agent, issued under the GNU Public License. The software allows Microsoft Agent applications to work on Windows 7 and above with support for existing Microsoft Agent characters, and Microsoft Office Assistant characters.   

The platform supports a scripting language, which can be used to generate executable files and interactive web pages. Scripts can be highly adaptive, with characters having multiple options to respond to a given cue. Once the parameters are set for communicating with a character via onscreen text prompting or voice commands, the only limit is your imagination.

Lots of Character

A step up from the Microsoft Office Assistant, the Double Agent character set includes animated figures like Merlin and the Genie. Characters can be configured for various interactive and entertainment programs. These range from speaking clocks and story readers to talking web pages.

Most are the kind of tool you’d employ to keep the kids occupied for several minutes while you’re downstairs making tea.

But there are some serious applications beyond the cute cartoon characters and pop-up speech balloons.

Artificially Intelligent Tutors

Looks aside, an agent is merely a computer program that has a specific plan of action defined in a limited domain and behaviour space.

As humans, we unconsciously and automatically interact with computers using social norms. Embodied conversational agents (ECAs) capitalise on this by presenting themselves to the learner as human-like characters capable of conversation with appropriate speech, gestures, and facial expressions.

ECA systems tend to be costly to build – hence the prevalence of proprietary software, like Double Agent. The best ones combine elements of gaming technology with low-level libraries like OpenGL.

Game engines usually have support for basic character animations, but may lack native lip-synch and fine animation controls. Typically, game engines come with a complex application programming interface (API) that has many features not immediately relevant for educational purposes, e.g. bullet/explosion physics, or first-person shooter perspectives.

By contrast, low-level libraries usually don’t have such irrelevant complexities. But they do require designing the artificial intelligence (AI) from the ground up.

A Girl Named Charlie

The turning point in Agent evolution came in the late 1990s, when developer Will Langstroth was commissioned to design an electronic tutor by South Africa’s University of The Western Cape.

Langstroth created virtual post-graduate assistant Dr Gisela Charlie, a pert, twenty-something redhead. Charlie was modelled on a real person, using motion-capture techniques now standard in Hollywood’s animated feature films.

The character boasted a comprehensive set of gestures and expressions – including smiles, frowns, yelling, yawning, and studying her fingernails when a course student takes a long time to give a response.

And no speech balloons.

The intention was to create a dialogue with an actual person – albeit a digital one.

Charlie served as the university’s Geographical Information Systems Electronic Lecturing Agent (GISELA – geddit?), for several years. After this, creator Langstroth released her into the wild as a free-to-download Microsoft Agent Character for personal computers.

Using natural speech based on Microsoft’s UK English Text-to-Speech Engine, Charlie went on to become (amongst other things) the digital spokesperson for a Europe-based chain of dance studios.

Talking Heads, and Such

Independent programmers and commercial enterprises have since built on this work. Photo-realistic 3D animated characters are now available free, or to buy, from numerous web outlets. Notable among these are the Microsoft Agent Ring, Guile 3D, and CodeBaby. Each proprietary system has its own scripting language, governing character animation and response.

Not just talking heads, either. Characters range from faces to full-body figures, and cover all ages, genders, and ethnicities.

Establishing Roles

In an Intelligent Learning Environment animated Agents can be used as cognitive tools, to support student learning. They may assume the role of virtual instructor, or that of a peer for the learner – an interactive course companion. The characters can provide sophisticated, real-time advice for solving problems, and a strong visual or personal appeal to make learning more engaging and effective.

Designing the Conversation

The main goal of an animated pedagogical (i.e., teaching) agent is to perform a given tutoring function for the student’s benefit. In a multi-agent environment, different agents may be used for different tasks.

For educational applications, the aim should be to choose a character that’s engaging – and one to whom your target audience can feel an instant connection.

Don’t be overwhelmed by the scripting requirements; the language is merely there as a framework, within which you craft a realistic conversation with the characters. The appearance and mannerisms of the Agent him/her/itself should help in defining the character’s role within the course structure you set up and establish a consistent “personality” which will govern its responses throughout.