In week 7 of our Enhanced Design Tools course, guest lecturer Dale Patterson gave a talk about data visualisation and artificial intelligence. This article is a summary of some of the concepts he discussed.
Data is all around us, and there is huge potential for designers to help people make sense of this sea of information. ‘Big data’ refers to the constant stream of data exchanged via the internet – it is big because of its high volume, velocity and variety. Big data comes from mobile devices (over 5 billion of them), internet-connected devices and computer systems.
Information visualisation can help people to understand data and get insights from it. Contemporary visualisation techniques enable interactive exploration of information and real-time analysis using big data. Designers have a key role to play by applying their visual skills and design knowledge to the visual communication of this data.
Artificial intelligence (AI) is a growing area of science that has the potential to help sort through and classify data, reduce manual processing and aid in visualisation of data. However AI cannot really understand data – it takes humans to do that. Designers should be aware of the potential and limitations of developments in AI.
There are three types of AI systems:
- Heuristics or rule-based: The system is assigned specific rules to follow in specific scenarios. Examples of these include systems used for direct mail, based on rules such as demographics and purchase history.
- Brute force: The system analyses every possibility, following a decision tree. For example a system playing a game like chess can analyse every possible move in order to identify the best approach. A limitation of this approach is that the system is only expert in one thing.
- Neural networks: The system learns how to classify information and recognise patterns based on data, similar to the way the brain processes and stores information by connecting neurons and transmitting messages. This is often called deep learning, which refers to the depth of layers in the network. For example, a system can be trained to recognise a specific pattern (such as an image of a face) if given enough data, and can improve its learning in response to continual data feedback. A limitation is the system needs a lot of data to learn, and learning how to do one thing (like identify dog images) doesn’t enable it to learn another thing (like identify cat images).
AI is good for limited system tasks and classification tasks. However it is less effective at making associations between concepts or understanding the implications of these. In other words, AI can help classify and sort data, but is not capable of assigning meaning to it.
This is where designers come in. Combining good visual and interaction design skills with the power of big data and AI is an area of real potential for designers. Coincidentally, this is discussed in a report I wrote for our Emerging Futures course: Visualising Big Data.