| A.I. | Human | Unsure | Droid | Robot | Error | Total | Turing Pct | Right Action |
| Alex | 7 | 3 | 4 | 32 | 7 | 53 | 15.2 | 6.6 |
| Roberta | 29 | 12 | 49 | 210 | 12 | 312 | 9.7 | 25.0 |
| George | 3 | 1 | 1 | 16 | 13 | 34 | 14.3 | 1.6 |
| Jabberwacky | 137 | 65 | 113 | 525 | 34 | 874 | 16.3 | 24.7 |
| Landru | 103 | 62 | 100 | 522 | 12 | 799 | 13.1 | 65.6 |
| Robby | 7 | 4 | 15 | 115 | 4 | 145 | 5.0 | 35.3 |
| Totals: | 286 | 147 | 282 | 1420 | 82 | 2,217 | 13.4 | 26.0 |
The Turing Test Simulator conducts a poll after each *successful* testing session. Your style and quality of conversation are properly noted. The answers to the poll questions are tabulated in the matrix above. Errors are assigned whenever the poll result is "unresponsive". This can occur because of Java compatibility issues and other technical problems, or it may represent the human participant's opinion that the computer was unresponsive. In either case this means that no conversation took place, so these responses are excluded from the Turing Percentage calculation.
The Turing Percentage divides the number of "human" responses by the sum of the first four columns. This number represents the percentage of tests in which the computer passed for human.
Turing's prediction: "in about fifty years' time [by the year 2000] it will be possible to program computers ... to make them play the imitation game so well that an average interrogator will have no more than 70 per cent. chance of making the correct identification after five minutes of questioning." (Turing 1950, p.442).
Turing's prediction would translate to a Turing Percentage of 30% as we are calculating it here.
The Right Action index is computed by dividing the sum of the first four columns by the "error" column. It is intended to give some idea how often the computer programs are giving plausible answers and are operating properly. Currently the error counts are very high, which may be a testament to the chat bots' reliability, or could be more indicative of variance in web browsers' Java implementations.
The Turing Hub 2006 begins June 1st.
The Turing Hub software is being revised and a new site is being created. The "switchboard" aspect of the automated system is being rewritten as a Java servlet. If you took part in the 2005 Turing hub tests, and would like to leave comments about how we might improve things, please contact us.
Turing Hub Interface Test: A prototype of the simulated "chat room" may be tried here. Java JVM required.
Turing Hub Subscribers' Bots: