Systems and methods for hybrid swarm intelligence
Systems and methods for real-time collaborative computing and collective intelligence are disclosed. A hybrid swarm intelligence system includes a central collaboration server, a plurality of computing devices in communication with the central server, and an agent application in communication with the central server. In response to information sent from the central server during a group collaboration session, user input is sent to the central server via the computing devices, and machine input is given to the server via the agent application, which determines input based on rules, additional data, and/or machine learning techniques. The central server uses the user input and the machine input to repeatedly provide feedback to the agent application and users during the group collaboration session.
1. A hybrid intelligence system for enabling human participants and machine agents to converge on an answer to a question together as a real-time dynamic system during a group collaboration session, the system comprising:
a central server in networked communication with a plurality of computing devices;
a software-controlled pointer for selecting the answer from a set of answer choices, the selection based at least in part upon a simulated spatial location of the software-controlled pointer in a simulated software environment;
a plurality of simulated target locations in the simulated software environment, each of the plurality of simulated target locations being associated with one answer choice of the set of answer choices;
a plurality of time-varying user input values, wherein each computing device is configured to repeatedly determine at least one user input value based on user input and repeatedly send the at least one user input value to the central server, wherein each time-varying user input value represents a user-directed motion of the software-controlled pointer with respect to the plurality of target location across a plurality of time steps during the group collaboration session; and
at least one time-varying machine agent value received by the central server, the at least one time-varying machine agent value representing an agent-directed motion of the software-controlled pointer with respect to the plurality of simulated target locations across the plurality of time steps, wherein the at least one time-varying machine agent value is determined using at least one pre-determined rule, wherein the central server is configured to update the simulated spatial location of the software-controlled pointer with respect to the plurality of simulated target locations in the simulated software environment based at least in part upon the plurality of time-varying user input values and the at least one time-varying machine agent value, resulting in a collaborative motion of the software-controlled pointer, whereby the answer is determined during the group collaboration session based at least on part upon an elapsed time and a relative location of the software-controlled pointer to at least one simulated target location.
2. The hybrid intelligence system of claim 1 , wherein each at least one time-varying user input value comprises, at least in part, a repeatedly updated user intent vector with a magnitude and a direction.
3. The hybrid intelligence system of claim 1 , wherein the at least one time-varying machine agent value comprises at least in part a repeatedly updated agent intent vector with a magnitude and a direction.
4. The hybrid intelligence system of claim 1 , wherein the at least one time-varying machine agent value is determined by a machine forecast system including forecast software configured to generate the at least one time-varying machine agent value in response to a machine model for answering questions, the machine model including the at least one pre-determined rule, and generate a repeatedly updated coordinate representing the location of the software-controlled pointer in the simulated software environment.
5. The hybrid intelligence system of claim 4 , wherein the machine model includes at least one percent likelihood associated with each of the plurality of answer choices.
6. The hybrid intelligence system of claim 4 , wherein the machine model includes a drift diffusion model.
7. The hybrid intelligence system of claim 4 , wherein the machine model includes a heuristic algorithm that starts with a preferred answer choice from the set of answer choices as a desired answer and switches to an alternate answer choice from the set of answer choices based at least in part upon how much time has elapsed and the collaborative motion of the software-controlled pointer over time during the group collaboration session.
8. The hybrid intelligence system of claim 7 wherein the at least one time-varying machine agent value indicates the agent-directed motion of the software-controlled pointer towards the preferred answer choice at a first moment in time and then switches the agent-directed motion of the software-controlled pointer towards the alternate answer choice at a second moment in time.
9. The hybrid intelligence system of claim 4 , wherein the machine model includes a heuristic model that starts with a preferred answer choice from the set of answer choices as a desired answer and switches to a second answer choice of the set of answer choices based at least in part upon the collaborative motion of the software-controlled pointer over an elapsed time and then switches to a third answer choice from the set of answer choices based at least in part upon and the collaborative motion of the software-controlled pointer over an additional elapsed time.
10. The hybrid intelligence system of claim 9 , wherein the at least one time-varying machine agent value indicates the agent-directed motion of the software-controlled pointer towards the preferred answer choice at a first moment in time and then switches to indicating the agent-directed motion of the software-controlled pointer towards the second answer choice at a second moment in time and then switches to indicating the agent-directed motion of the software-controlled pointer towards the third answer choice at a third moment in time.
11. The hybrid intelligence system of claim 4 , wherein the machine model uses as input information regarding the plurality of time-varying user input values.
12. The hybrid intelligence system of claim 11 , wherein the machine model is responsive at least in part to a distribution of users pulling for each of a plurality of different answer choices, each from the set of answer choices, over time.
13. The hybrid intelligence system of claim 4 , wherein the machine model is derived at least in part by performing machine learning upon a plurality of data sets collected when a group of human users control the software-controlled pointer when answering a plurality of previous questions of similar structure to the question.
14. The hybrid intelligence system of claim 4 , wherein the machine model uses an allowed time left for answering the question as a parameter for determining the at least one time-varying machine agent value.
15. A method for enabling human participants and machine agents to converge on an answer to a question together as a real-time dynamic system during a group collaboration session comprising:
sending of initial session information including a set of answer choices, each answer choice associated with one of a set of simulated target locations in a simulated software environment, from a central server to a plurality of computing devices in communication with the central server;
sending, repeatedly in real-time during the group collaboration session, of a user input value from each computing device to the server in response to user input, wherein each time-varying user input value represents a user-directed motion of the software-controlled pointer with respect to the plurality of simulated target locations across a plurality of time steps during the group collaboration session;
sending to the central server, repeatedly in real-time during the group collaboration session, of at least one time-varying machine agent value from an agent application in communication with the central server, the at least one time-varying machine agent value representing an agent-directed motion of the software-controlled pointer with respect to the plurality of simulated target locations across the plurality of time steps, wherein the at least one machine agent value is determined using at least one pre-determined rule; and
updating by the central server, repeatedly in real-time during the group collaboration session, the simulated spatial location of the software-controlled pointer with respect to the plurality of simulated target locations in the simulated software environment based at least in part upon the plurality of time-varying user input values and the at least one time-varying machine agent value, resulting in a collaborative motion of the software-controlled pointer, whereby the answer is determined during the group collaboration session based at least on part upon an elapsed time and a relative location of the software-controlled pointer to at least one simulated target location.
16. The method of claim 15 , wherein each time-varying user input value comprises, at least in part, a repeatedly updated user intent vector with a magnitude and a direction.
17. The method of claim 15 , wherein the at least one time-varying machine agent value comprises at least in part a repeatedly updated agent intent vector with a magnitude and a direction.
18. The method of claim 15 , wherein the at least one time-varying machine agent value is determined by a machine forecast system including forecast software configured to generate the at least one time-varying machine agent value in response to a machine model for answering questions, the machine model including the at least one pre-determined rule, and generate a repeatedly updated coordinate representing the location of the software-controlled pointer in the simulated software environment.
19. The method of claim 18 , wherein the machine model includes at least one percent likelihood associated with each of the plurality of answer choices.
20. The method of claim 18 , wherein the machine model includes a drift diffusion model.
21. The method of claim 18 , wherein the machine model includes a heuristic algorithm that starts with a preferred answer choice from the set of answer choices as a desired answer and switches to an alternate answer choice from the set of answer choices based at least in part upon how much time has elapsed and the motion of the software-controlled pointer over time during the group collaboration session.
22. The method of claim 21 , wherein the at least one time-varying machine agent value indicates the agent-directed motion of the software-controlled pointer towards the preferred answer choice at a first moment in time and then switches to indicating the agent-directed motion of the software-controlled pointer to the alternate answer choice at a second moment in time.
23. The method of claim 18 , wherein the machine model includes a heuristic model that starts with a preferred answer choice from the set of answer choices as a desired answer and switches to a second answer choice from the set of answer choices based at least in part upon the collaborative motion of the software-controlled pointer over an elapsed time and then switches to a third answer choice from the set of answer choices based at least in part upon and the collaborative motion of the software-controlled pointer over an additional elapsed time.
24. The method of claim 23 , wherein the at least one time-varying machine agent value indicates the agent-directed motion of the software-controlled pointer towards the preferred answer choice at a first moment in time and then switches to indicating the agent-directed motion of the software-controlled pointer towards the second answer choice at a second moment in time and then switches to indicating the agent-directed motion of the software-controlled pointer towards the third answer choice at a third moment in time.
25. The method of claim 18 , wherein the machine model uses as input information regarding at least one time-varying user input value.
26. The method of claim 25 , wherein the machine model is responsive at least in part to a distribution of users pulling for each of a plurality of different answer choices over time.
27. The method of claim 18 , wherein the machine model is derived at least in part by performing machine learning upon a plurality of data sets collected when a group of human users control the software-controlled pointer when answering a plurality of previous questions of similar structure to the question.
28. The method of claim 18 , wherein the machine model uses an allowed time left for answering the question as a parameter for determining the at least one time-varying machine agent value.
29. A hybrid machine intelligence method using a machine intelligence system and human scouts, the method comprising:
providing a central forecasting server in communication with a plurality of computing devices, each of the plurality of computing devices running a scouting application for use by a human user;
sending a forecasting query from the central forecasting server to each of the plurality of computing devices, the forecasting query indicating a forecasting question to be researched by the human user of the computing device by searching the internet;
enabling the human user of each computing device to search the internet in response to the forecasting query received from the central forecasting server while recording tracking data as the human user accesses web-based documents responsive to the forecasting query;
enabling the human user of each computing device to provide evaluation data regarding the web-based documents accessed by the user in response to the forecasting query, the evaluation data including an indication of at least one of importance, bent, and validity of each document accessed;
sending to the central forecasting server, from each computing device, scouting data regarding the web-based documents accessed by the human user including the tracking data and the evaluation data; and
analyzing, by the central forecasting server, the scouting data from the plurality of human users in response to the forecasting query to produce a machine forecast, said machine forecast providing a set of probabilities regarding the likelihood of each of a plurality of possible outcomes related to the forecasting query.
30. The hybrid machine intelligence method of claim 29 , wherein the steps of searching, tracking, and providing evaluation data are performed by an instrumented web browser on each of the plurality of computing devices that is configured to send data to the central forecasting server.