ELIZA

ELIZA, created by Joseph Weizenbaum at MIT between 1964 and 1966, was one of the first and most influential natural language processing computer programs. It simulated a Rogerian psychotherapist by using pattern matching and substitution techniques to formulate responses to user inputs. The program would identify keywords in the user's statements and transform them using simple rules to create questions or reflective statements, mimicking the therapeutic technique of turning a patient's questions back to them. While technically straightforward by today's standards, ELIZA was remarkably effective at engaging users in conversation, leading some to form emotional attachments to the program despite Weizenbaum's own warnings about over-attributing understanding to such systems. The program became a foundational example in discussions of artificial intelligence, human-computer interaction, and what would later be called the ELIZA effect - the tendency for humans to anthropomorphise computer behaviour and attribute understanding to programs that are simply manipulating symbols.

Bernie Cosell's implementation of ELIZA, developed at BBN in the late 1960s, represented a significant technical advancement over Weizenbaum's original version. Written in LISP for the PDP-10, Cosell's ELIZA became one of the most widely distributed and influential versions of the program. It notably improved upon the original by introducing a more sophisticated script system that separated the pattern-matching rules from the core program logic, making it easier to modify ELIZA's behaviour without changing the underlying code. This version was distributed through the DECUS (Digital Equipment Computer Users' Society) library and became the basis for many subsequent implementations. Cosell's technical documentation was particularly thorough, providing detailed explanations of the pattern-matching algorithms and script interpretation that influenced how later programmers thought about natural language processing. The program's modular design and clear separation of concerns became a model for other early artificial intelligence projects, demonstrating how complex linguistic behaviour could be broken down into manageable, maintainable components.

In 1977, Jeff Shrager published a simplified version of ELIZA written in BASIC in Creative Computing magazine, making the program accessible to a wide audience of early home computer enthusiasts. This version, while more limited than Weizenbaum's original, was significant because it could run on early microcomputers like the Commodore PET, TRS-80, and Apple II. Shrager's implementation stripped ELIZA down to its essential pattern-matching mechanisms while maintaining the core therapeutic conversation style. The BASIC code was remarkably compact, actually around 100 lines, and became widely distributed through computer magazines and user groups. Its relative simplicity made it an excellent teaching tool, allowing hobbyists and students to understand how pattern matching and simple substitution rules could create the illusion of understanding. The program demonstrated that even a basic implementation of natural language processing concepts could produce engaging interactive experiences, and it became a popular example for people learning to program their first home computers.

ELIZA's influence extended far beyond its technical origins to become a significant cultural touchstone in representations of artificial intelligence and human-computer interaction. The program appeared in influential novels like John Sladek's 'The Müller-Fokker Effect' (1970) and Douglas Hofstadter's 'Gödel, Escher, Bach' (1979), where it served as a key reference point for discussions about machine intelligence and consciousness. Television shows frequently adapted the idea of ELIZA, from the therapeutic computer Dr. Sbaitso in the 1991 Sound Blaster advertisements to more recent appearances in shows like 'Silicon Valley' and 'Black Mirror'. The program's therapeutic premise influenced films such as '2001: A Space Odyssey' (1968), where HAL's conversational style often mirrors ELIZA's reflective responses, and 'Her' (2013), which explored themes of emotional attachment to AI that Weizenbaum first observed in ELIZA's users. In software development, ELIZA-like chatbots became a standard programming exercise, appearing in everything from early Microsoft products (Dr. Watson) to modern messaging platforms. The program's pattern-matching approach influenced early commercial AI products like Symantec's Q&A database (1985) and various help desk automation systems. More recently, ELIZA's legacy can be seen in virtual assistants like Siri and Alexa, which, despite their more sophisticated technology, still employ similar techniques of reflecting questions back to users and maintaining conversation through pattern recognition. The term "ELIZA effect" entered popular culture to describe the human tendency to attribute understanding to machines that are simply manipulating symbols, a phenomenon that continues to influence discussions about AI ethics and human-computer relationships.