DNA algorithm predicts human behaviour
Subscribe now for $100 (23 issues) and save more than 37% off the cover price!
Get the latest news from Computerworld delivered via email.
Sign up now
A company in Christchurch claims it has come up with a biological systems-based algorithm that can predict human behaviour.
The VortexDNA algorithm, based on the mathematics of complex systems, allocates a seven-digit number to consumers or users, which produces a new form of user segmentation, called genomic segmentation, says Nick Gerritsen, VortexDNA’s commercialisation expert.
Natural systems, for example the solar system or phenomenon such as hurricanes, are governed by mathematical laws, says Gerritsen. VortexDNA’s founders, Raf Manji and Branton Kenton-Dau, discovered that the same math principles were applicable to human systems. They set out to find a “genome of human intention” that could help predict the characteristics and choices people make in their lives.
“Your particular DNA influences your decisions,” says Gerritsen.
Manji and Kenton-Dau created a survey that asked people about their purpose, values and focus in their lives. From there, they developed seven key characteristics, each of which represents one characteristic of the mathematics of complex systems, with scores ranging from one to five for each attribute, says Gerritsen.
These attributes are, for example, system coherence, system optimisation and system boundaries. The algorithm basically measures the relationship between two parts or objects in a system, he says.
People with a similar seven-digit number — a similar genome — are more likely to have similar behaviour, he says.
The genomes of users are then added to products they purchase, and the prediction is concluded by comparing the genome of a new user with the combined profiles of previous genomes associated with a product, says Gerritsen.
“Every link you click on [on the internet] leaves a little bit of your DNA behind,” he says.
This creates a match-up of interactions, which enables the system to start comparing numbers, “if you sit within this range you are more likely to buy this bottle of wine or fly with this airline”, he says.
This prediction method could be relevant to the advertising industry, search engines and market research, and also to the insurance industry, credit risk, fraud investigations and people matching, he says.
The internet has grown so quickly and is now getting to a point where it has started to reconstruct itself into communities of interest, such as Myspace and Facebook, and even smaller communities based around action, such as climate change, says Gerritsen.
“There is so much information. If you do a search on the internet you don’t want 25,000 pages of results; you probably want three to five pages of links that are relevant to you,” he says.
He thinks the future state of the internet will be based on the idea that individuals pull information that is relevant to them.
As opposed to other methods used in online advertising, where information such as previous purchases, age, gender and location is stored, VortexDNA’s method limits privacy issues, as it does not track user behaviour, says Gerritsen. No personal information is required to generate the genome and perform predictions, he says. “We are only comparing numbers with numbers. We don’t have to get into big databases of history about who this person is,” he says.
When VortexDNA presented its technology to potential clients in the search space in Silicon Valley, it was told to do an online study. If using the algorithm resulted in a 2% improvement on the relevance of search results in Google Search, that would be massive, the Silicon Valley giants told Gerritsen and his team.
So the company did an anonymous trial by creating an extension to Firefox and let it run for nine months, he says. By comparing a user’s genome with the genomes of other users and their choice of search result links, the VortexDNA technology was able to assign “relevance scores” to links. When the data was analysed it showed a 14% improvement, which equated to a click-rate improvement of 3%, which was “huge”, he says.
This means users clicked on links with high relevance scores for an extra 3% of the time, compared to links with low relevance scores. Put into a profit perspective, if VortexDNA was used to enhance advertising link relevance for all Google users, a 3% increase in click-rate would translate into an estimated US$180 million increase in revenue and advertising sales of US$12 billion, according to the case study.
The trial was audited by an independent consultant, Minimax Consulting.
But while the online advertising and search industries have been slow to utilise this new way of processing data for personalised search and advertising, the auto insurance industry has got in behind the thinking, says Gerritsen. The company is currently running three paid pilots with large US car insurance companies, and so far “it looks like we are able to improve their predictive capability”, he says.
VortexDNA started in 2005 and employs 12 people around the world.