Meteosam Source: Screenshot, by the author Click on the image to enlarge it

Sam is a virtual weatherman, a synthetic meteorologist that, in the usual style of its human homologues, presents weather forecasts the TV way.

MeteoSam is actually a set of software programs that enable Sam to generate a forecast and construct the sentences he will say, the voice synthesis and the generation of bodily movements, from the data entered into a database by professional meteorologists. The facial expressions correspond to the sentences he says and are also generated automatically by the computer.

Moreover it's a multi-platform system that adapts itself automatically not only to the data transmission format of each one of the three different platforms it covers (Television, Web and Mobile Phone) but it also adapts Sam's graphical representation, movement and explanations to them. Sam has been developed within the framework of an innovation project by a consortium led by Activa Multimedia in Spain.

Technologically it's an advanced system that, unlike other systems available on the market, it's not only a talking head but it moves hands and the rest of its body.  The weather forecast it provides is complemented with the iconic traditional way in meteorological spaces with icons depicting sun, rain, etc.

Sam is one amongst the many virtual characters that have been populating cyberspace in the last decades. We won't speak here about the virtual movie stars or the main characters of videogames whose main goal is entertainment and whose natural space is precisely this. There's no doubt, given the ubiquity they show, that this is an outstanding success of virtual characters, widely accepted by the public.

We are more interested by those, like SAM, that constitute a platform for the transmission of information more than a way of entertainment. The conversion of computerised information in vocal expression and gestures is by no means an easy task, despite the noticeable advances that have been produced with Sam and other similar characters. Moreover Sam and colleagues have to face the doubt of whether the information they transmit is easier to understand than the same transmitted through some of the well known and established ways.

Meteosam: Today's forecast Source: Screenshot, by the author Click on the image to enlarge it	Meteosam: Looking at temperatures Source: Screenshot, by the author Click on the image to enlarge it

According to DA Group, an English group specialized in the creation of virtual characters and all its associated software "Virtual people,..., offer a far more accessible way to interact with the computers and machines that dominate everyday life. The challenge is to create characters that are as lifelike and as responsive as another human being whilst capable of accessing, assimilating and filtering vast amounts of computer knowledge".

The utility of such characters is centred on mediating with human beings accessing the services they provide. Among them we can count:  

• citizen information provided by public governmental organisations
  
  
• customer service, substituting automatic phone services 
  
  
• learning medical diagnostic techniques in medicine, virtual pacients 
  
  
• e-learning mentorship in online courses, answering frequently asked questions 
  
  
• personalised news presentation, like Sam

In the end all of them represent a kind of [anthropomorphic]user interface, that isn't based on the typical widgets and icons, although it can be combined with them. 

An example of application of services for the citizen is tourist assistant a demo (also from the DA group) that provides tourist information for the city of London, indicating the main attractions of the City and the best itineraries by underground, bus and taxi. 

 Probably the easiest to assimilate virtual characters at this moment are those that accomplish similar functions as their human counterparts. This includes TV characters, weather men and presenters, all of them with a certain entertainment facet.  There's no doubt about the utility of a personal agent accessible from different media 24 hours a day from every place in the world. If instead of behaving mechanically with a very restricted intelligence level it can provide us with useful information in an intelligible way, its utility is unbeatable.

Here's where avatars have their Achilles'heel. The intelligence and possibilities of many of them are still far too limited, as examples like the unbearable Clippo and its colleagues acting as help assistants in MS Office vouch for. It's difficult to say what is the point in having help lists accompanied by a supposedly funny (yet of scarce intelligence) virtual character.

Ananova: Aspect offered by the virtual character when the web was launched.  Source: Google images

Many anthropomorphic agents experiences have been produced trying in one way or the other to introduce them into our world. An example is Ananova a company that provides news online that launched a virtual character able to read news and present it "live". In 2000 Ananova was bought by Orange and at least since 2002 it no longer uses its femenine avatar, although it's claimed to be still in development.

On the other hand some recent experiments with advanced avatars show that they can be very persuasive, just by using very simple techniques, like copying subtly the gestures of its interlocutor with some delay (what psychologists call chameleon effect). A study performed by researches of the Virtual Human Interaction Laboratory at Stanford University showed that the students that they tried to persuade of the need to permanently carry their ID with them while at the University by subtly imitating their gestures found the avatar more convincing and pleasant than those that they tried to persuade without the copying gestures technique.

The implications of this technique are somewhat disturbing and probably deserve further discussion in another place.

It's still a question of debate whether an avatar can be more useful that the information presented in textual or graphical form. The same occurs with many 3D visualisation experiments that usually are less operative that their 2D counterparts, specially if the application doesn't intrinsically require a 3D environment.

So while we wait for an increase in the artificial intelligence  of these agents and we contemplate how graphic representation and movement and facial expression modellisation advance at a steady pace, we can connect to MeteoSam to see what the weather will be like tomorrow.

Links of this issue: