Micro – Why not human and AI (loooonger post) 1a.

Its very complex and I’m not sure – so here goes.  This post makes a lot of reference to Second Life so if you have not experienced Second Life, heard of an avatar or virtual worlds it will probably be somewhat strange.

Firstly – I have watched quite a bit of sci-fi but never read a sci-fi book so – looking at what the American Association for Artificial Intelligence has to say (and especially the John McCarthy article) it is intelligence that is computational, possibly not observable, now used to study pattern recognition, complex decision making, the use of natural language – and there are many more definitions. So it includes complex decision making so surely this fits into Micro – maybe !

Role playing and identity in a virtual world – does this affect decision making ?

Last year at Surrey, a student – Anna Vartapetiance-Salmasi completed her dissertation on Machine Ethics using Second Life as an environment. She has presented at various conferences including EthiComp, IEEE Virtual Worlds and also at the National Workshop at  Coventry (for some reason there now seems to be a login to watch the video… investigating…, but the ppt is fine)

Earlier last year she asked me to take part in the questionnaire – you can also take it – if you go to the doors of EthiCasino on Surrey Island. My avatar – Esme Delcon  – in her results on the slides, appeared in the high-risk category. I have not spent a great deal of time in casinos in my real life or gambling of any kind – I find it very boring so have very little knowledge of the ‘how’ and ‘rules’ but I would describe myself as a risk-taker from time to time. I have also not looked at the studies Anna referenced so have very little understanding of machine ethics, especially the technical  – programming languages & how the logic of the languages call different types of actions etc

My avatar – I deliberately chose in Second Life to keep my avatar relatively close to my real life physical appearance  and have not made a decision to play a different role to my real self, as far as I am aware. Professor Maggi Savin-Baden (Coventry Uni) has written a ton of stuff on identity in Second Life. Some people do actively take on a different role – there is probably some research on avatar identity and decision making, but I can’t think off top of my head (David White – Habitat – possibly?). Either way – it is still a human making conscious choices, possibly different aspects ? So I can’t claim that my avatar is a high-risk gambler operating separately from my conscious (or subconscious) self?

Anna’s work is amazing but her study was in Second Life with 50 avatars and taking the limitations of Second Life and how much complex programming can be used to simulate a real life environment. Also would I answer the questionnaire in the same way every time – especially in real life if I entered through casino doors either on or offline, how many other factors also affect how I answer questions – the limitations with data analysis exist too?

Artificially Intelligent in Second Life

There are now bots in Second Life which are not operated by a human. We have started some research in this area at Surrey (please contact Lee Gillam for more info – due to time & resource constraints we were unable to produce anything for PREVIEW in time but it is ongoing research regarding avatar and information driven problem based learning scenarios). PIVOTE is an open source output from that project from St George’s – you can find it via PIVOTE references on the PREVIEW blog. The bots in the PREVIEW project are mostly artificially intelligent bots in medical or healthcare simulations which respond to avatar questions and perform tasks.The report has some info about whether artificially intelligent bots can learn or not but is a more general review of the project.

As well as bots – you can also place scripting on objects (I guess you would call these bots too but they might not have an avatar appearance and/or be visible) in Second Life which can record information about avatar or environmental activity in that area, respond to questions  too.

Daden have a huge range of experience in creating automated avatars in Second Life:

But this post is just skimming over everything… So far the following – the role of identity and whether this is at a human conscious or subconscious level, responding to an avatar and the programming providing results based on the input of a human operated avatar, how an avatar – human or non-human behaves in a virtual environment and if any of the conditions of the environment or conversation taking place change – does any artificial intelligence in the form of avatars or the environment adapt / learn / do something differently as a result of receiving and processing different information?

A much more detailed look at artificial intelligence which was written way before (2000)  Second Life became popular – covers autonomy, human cognition and an overview of virtual reality and virtual environments. These are a lot of different areas and there has been considerable research looking into virtual reality but within Second Life, there is not a large amount of research (that is known about ?) so it is very difficult to draw conclusions about artificial intelligence and decision making due to the huge number of different areas to look at.

I find this similar to behavioural economics perspectives on how we make decisions (future post) because there are so many different factors which can affect decision making and evidence of how a human and a computer/machine interact, in addition to responses in software intelligence may all be part of how a decision is made – not just the ethical, rational perspectives but the physical/biological responses at that moment from the human and the resulting interaction with the computer/machine.  AI and economics is also not a new concept and is gaining popularity. This is why at this moment, I find it difficult to try and understand economics at a personal, micro level if human + AI decision making is involved – but I do think this may be easier at a macro level.

1b will look at decision making in more detail and AI in non 3D virtual environments in relation to economics.