To make things easier, we've compiled a jargon-busting glossary of straightforward but detailed definitions for all the must-know terms in algorithmic underwriting.
An algorithm is essentially a set of instructions that helps a computer to complete any given task or calculation. In algorithmic underwriting, algorithms are instructions that tell computers how to underwrite on their own and improve their decision making over time.
Application Programming Interface (API)
An API, or an application programming interface, is a type of intermediary that enables two separate applications to communicate with one another. In algorithmic underwriting, an API can be used to create a connection between an underwriting algorithm, a broker and/or data sources to obtain real-time pricing data and improve the efficiency and accuracy of the algorithm's output.
In simple terms, APIs connect disparate systems and data sources to improve efficiency and increase revenue.
Artificial Intelligence (AI)
Artificial intelligence (AI) is the field of computer systems and applications that can complete tasks that traditionally require human intelligence. The ultimate goal of AI is to create computer programmes that can 'think' and behave like humans, including carrying out decision-making and problem-solving tasks.
AI is an umbrella term that incorporates many different elements, such as machine learning and deep learning, both of which are complex functions of AI-capable systems. While AI concerns the ability of computers to 'think', machine learning focuses more specifically on the ability of computers to 'learn'.
Big data refers to extremely large data sets that usually grow extremely quickly - faster than any human could attempt to keep up with. Big data can usually only be sorted, analysed, and queried thanks to technology like artificial intelligence.
Big data in the insurance industry can be collected from a variety of sources, including historic customer data, public records, and telematics. With thousands of data points to work with, big data helps AI algorithms to identify new patterns and risk indicators all the time.
Data scientists are qualified professionals responsible for collecting, categorising, and analysing sets of data. They combine knowledge from a number of traditional fields including mathematics, statistics, and science and use data science technologies to develop algorithms and carry out predictive modelling.
In algorithmic underwriting, data scientists may develop the algorithms and models that underpin the technologies that underwriters use to automate tasks.
Deep learning (DL)
Deep learning (DL) is a subfield of machine learning, which in turn is a subset of artificial intelligence, that concerns artificial neural networks or algorithms that mimic the structure and processes in the brain.
Deep learning enables computers to make predictions from large data sets with high accuracy by working with complex neural networks of three or more layers. Deep learning is responsible for some of the most advanced AI algorithms in algorithmic underwriting today.
Machine learning (ML)
Machine learning (ML) is a subset of artificial intelligence that uses data and algorithms to enable computers to 'learn' in a way that mimics human learning. Machine learning-enabled systems are capable of learning from user’s feedback, becoming more accurate and more efficient with every data point they analyse over time.
There are many branches within machine learning. Here are some of the most relevant to underwriting:
1. Natural language processing (NLP)
Natural language processing (NLP) is a branch of machine learning that focuses on programming machines to learn, process, and use human languages. Good examples of NLP in action are AI voice assistants like Alexa and Siri.
NLP enables computers to make sense of written text and carry out lingual tasks including translation, subject classification, and keyword extraction. In algorithmic underwriting, NLP might be used to help underwriting algorithms parse qualitative, text-based data and output reports.
2. Optical character recognition (OCR)
Optical character recognition (OCR) is a set of techniques that enables computer programmes to detect text in images.
For example, if a programme with OCR technology scans an image of a crash site when assessing a claim on car insurance, the programme may be able to gain additional information about the scene by parsing the text data on nearby speed limit signs, traffic signs, and number plates.
3. Robotic process automation (RPA)
Robotic process automation (RPA) is a type of programme aimed at assisting humans with manual tasks such as data entry. While machine learning is a data-driven technology, RPA is a process-driven technology.
RPA is useful in situations where computers are set to carry out repetitive, rules-based processes. RPA technology can automate some of the most time-consuming manual tasks in underwriting (e.g. rekeying of data into different systems) and save underwriters hours of valuable time.
Structured data is a type of data that is easy for machine learning algorithms to parse and understand. It's managed by structured query language (SQL) and usually stored in an SQL database. Examples of structured data in insurance include any table-like structures (e.g. spreadsheets representing a claimant's age, claims history, and other relevant data points).
Unstructured data is qualitative data that isn't easy for computers to readily process and analyse without advanced machine learning technology. It can be managed in NoSQL databases or preserved in its raw form as files within data lakes. Examples include free form documents (Word, PDFs), images, or sounds.
Machine learning algorithms can mine unstructured data in large volumes to extract important insight. For example, data relating to consumer behaviour and sentiment can in turn inform up-to-date market pricing and risk analysis outcomes.
For more information on how to apply these technologies to your insurance business, get in touch.