Classifications

• • G—PHYSICS
  • G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
  • G09B—EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS
  • G09B7/00—Electrically-operated teaching apparatus or devices working with questions and answers

Definitions

• the present invention relates to the following technical fields: distance learning, computer-mediated communication (CMC), forum systems and artificial intelligence.
• CMC computer-mediated communication
• a distance learning environment allows for carrying out learning activities among dispersed people by providing three main functions: course management, content organization and group communication.
• the present invention relates to the group communication functions; the paragraphs below briefly describe the background art on this matter.
• Group communication in distance learning environments, can be synchronous or asynchronous, depending on whether the learners are required to be connected at the same time (synchronous) or they can communicate in different times (asynchronous).
• Asynchronous group communication can be divided in two types: message-based and discussion-based.
• Message-based communication e-mail
• Discussion-based communication allows people to collectively debate over a set of proposed issues. Group discussions are carried out in the so-called “discussion forums", i.e. topic-driven virtual spaces where the participants can add questions, give answers and argue over their peers' opinions.
• Discussion support tools are generically known as "forum systems" and can be either integrated in a distance learning environment or developed a stand-alone system.
• the architecture of forum systems comprises two elements: (i) a database storing the participants' postings and (ii) a user interface allowing the participants to view the discussion (generally in a threaded format) and to place their respective postings.
• the COMPUTATIONAL METHOD FOR MEDIATING ASYNCHRONOUS GROUP DISCUSSIONS proposes an alternative way of managing group discussions, in which the discussion mediation is entirely algorithmic, thus allowing the method to be fully implemented by a computer program.
• the proposed solution consists in turning group discussions into "argumentative exercises", where the participants are intentionally stimulated to provide answers, to argue over their peers' ideas and to defend their own viewpoints.
• the proposed method combines artificial intelligence techniques with argumentative discussion to allow the design of a new category of forum systems - the "self-mediated (active) forum systems".
• the first advantage of the present method when compared to the currently existing forum systems, is the possibility of conducting group discussions among an unlimited number of participants (typically among tens or hundreds of participants) over several different issues simultaneously, with no human mediating effort. This relieves the professor/tutor from the time- consuming task of articulating the discussion and gives him/her additional time to perform more cognitive tasks in the discussion, such as clarifying concepts, validating polemical viewpoints or mediating conflicts of ideas.
• the present method allows group discussions to advance in an intentional manner, so as to optimize the discussion outcomes in terms of group interaction.
• the optimization of the discussion is done by (i) focusing the discussion on polemical viewpoints (rather than on common agreements), (ii) evenly distributing the participants over the proposed issues and (iii) stimulating the learners to take part of the discussion by sending them customized discussion forms.
• the present invention allows the design and implementation of "self- mediated forum systems", i.e. traditional forum systems added with a discussion mediation mechanism.
• the architecture of such a self-mediated forum system should be composed of the following elements: (i) a database storing the participants' postings, (ii) a user interface allowing the participants to act on the discussion, (iii) a mediation mechanism performing the tasks presented in this description and (iv) a scheduler/timer triggering the mediation mechanism according to a predefined schedule or to a periodic time interval (if a fully automatic mediation is intended).
• Such a self-mediated forum system can be designed either as a standalone software system or can be integrated to an existing distance learning environment. In the latter case, the mediation mechanism and the scheduler/timer can be added as background elements to the existing discussion forum of such an environment.
• the COMPUTATIONAL METHOD FOR MEDIATING ASYNCHRONOUS GROUP DISCUSSIONS consists in launching a set of issues for debate and redistributing the participants' answers and argumentations among the group to be collectively validated in successive discussion rounds (cycles). At each discussion cycle, the method detects disagreements and proposes new interactions among the group. The participants take part in the discussion by filling customized discussion forms, containing questions/answers to be answered/argued. Such discussion forms are algorithmically produced by the method and can be delivered to the participants through the Internet.
• FIGURE 1 showing an example of a discussion tree
• FIGURE 2 illustrating the overall flowchart of the method
• FIGURE 3 illustrating the flowchart of the NEW CYCLE procedure
• FIGURE 4 showing an example of discussion form (1 st cycle);
• FIGURE 5 showing an example of discussion form (after 1 st cycle);
• EQUATION 3 presenting the formula for calculating the re-launch score of a given node N, named RS(N).
• the discussion tree is made of a root node (1) connected to a set of "question nodes" (Q nodes) (2), where each Q-node can be linked to a number of "answer nodes” (ALT nodes) (3) and each answer node can be associated to a sub-tree of "argumentation nodes” (ARG nodes) (4).
• Each Q node has a textual content representing a question to be debated.
• Each ALT node has (i) a textual content representing an answer to the corresponding question and (ii) the identification of the participant who is the author of the answer.
• Each ARG node has (i) a refuting/supporting intention in respect to the corresponding parent node, (ii) a textual content representing the argument that holds the intention and (iii) the identification of the participant who is the author of the argumentation.
• an ALT node is a triple of the form ⁇ p, q, ⁇ >, where p represents the content of the answer, q represents the question to which the answer refers and i represents the author of the answer.
• an ARG node is a 4-tuple of the type ⁇ g, h, , ⁇ >, where g represents the text of the argument, h represents the answer/argumentation to which the argumentation refers, represents the intention to support/refute h and i represents the author of the argumentation.
• the four possible values of define four subtypes of ARG nodes, respectively ARG++, ARG+, ARG- and ARG—.
• Each node of the discussion tree represents a contribution (posting) of a given participant to the discussion; the relation between two nodes represents a specific peer-to-peer interaction.
• the essence of the method is to analyze the existing relations among the nodes and to create new interactions in order to advance the discussion in an intentional manner.
• the creation of a new interaction is done by "re-launching" an existing node, i.e. by creating a new node as a direct descendant (child) of an existing node and assigning the newly created node to a given target participant (see the RELAUNCH procedure for further details).
• the re-launch of a given node N to a target participant I results from applying an "assignment" of the type (N, I).
• the discussion is advanced by re-launching a number of nodes, based on the desired peer-to-peer interactions, which causes the current discussion tree ( ⁇ n ) to advance to the subsequent cycle ( ⁇ n+1 ). Successive discussion cycles are opened until the discussion cannot advance any further, as described below.
• the present method is composed of a sequence of procedures, as illustrated by the general flowchart shown in FIGURE 2.
• the procedures are represented by rectangles and the corresponding input/output resources are linked to the procedures by dotted lines.
• the items below detail each of the procedures, by describing them in textual form and, when necessary, by the corresponding algorithm.
• the method initiates with the PREPARATION procedure (7), which is responsible for creating the upper level of the discussion tree (i.e. root node and Q nodes) and distributing the set of questions (5) among the set of participants (6).
• the output of the PREPARATION procedure is a set of assignments of the type (Qi, Ij), where Qi is a proposed question and Ij is the target participant. This distribution of questions among the participants is done in such a way that each participant is assigned the same number of questions and the number of questions assigned to each participant does not exceed the maximum workload per cycle (WL max parameter).
• the PREPARATION procedure is defined as follows. Let Q be the list of questions and Q[l] ... Q[m] be the individual questions to be debated. Let I be the list of participants and ⁇ [l] ... ⁇ [n] be the individual participants of the discussion. Let i and j be two index variables, respectively used to refer to elements of Q and I. Let count be a variable used to count the number of questions already assigned to a given participant. Let WLmax be a parameter representing the maximum workload for each participant in a given discussion cycle, i.e. the maximum number of items in a discussion form.
• length(A) be a function that returns the number of elements of the A list and child- ⁇ st(N) be a function that returns a list of direct child nodes of a given node N.
• parent(N) and text(N) be properties of a node N, respectively specifying the parent node and the textual content of N.
• create-ne -qnode be a function that creates and returns an empty Q node.
• the RELAUNCH procedure (8) aggregates new nodes to the discussion tree according to a list of assignments of the type (N, I), where N is the node to be re-launched and I is the target participant.
• N the node to be re-launched
• I the target participant.
• the re-launch of N expands the tree, either in depth or in breadth, by a new node; in what concerns the group interaction, the re-launch of N creates a new peer-to-peer interaction between the author on N and the target participant I.
• the RELAUNCH procedure is defined as follows. Let assign-list be the list of assignments taken as input parameter.
• current-assign be an element of assign-list and current-assign [1] and current-assign [2] be respectively the current nods to be re-launched and its target participant.
• type (N) , parent (N) , author (N) , text (N) and w (N) be the properties that specify respectively the type, the parent- node, the author, the textual content and the argumentative intention of N.
• create-aitnode and create-argnode be two functions that respectively create an empty ALT node and an empty ARG node.
• the RELAUNCH procedure aggregates new nodes to the discussion tree and assigns them to the corresponding target participants. However, the content of such nodes (answers/argumentations) are still empty and must be filled by the respective target participants. In order to allow the participants to provide their answers and argumentations, the discussion is delivered to them in customized discussion forms, as described below.
• the DELIVERY procedure (9) builds and delivers customized discussion forms (14) that allow the participants to provide their respective answers and/or argumentations.
• the DELIVERY procedure retrieves the newly created nodes from the discussion tree and generates discussion forms, as follows. For each participant Ij of the discussion, a set Sj of all newly created nodes assigned to I is determined. Each node N in Sj is then converted into a "discussion form element" (30), more specifically an entry of the discussion form formatted in such a way that the whole path from N up to its corresponding Q- node appears in the form (to allow for visualization of the whole discussion thread) and the participant can enter the respective answer/argumentation (see examples in FIGURE 4 and FIGURE 5). Once all elements of Sj are converted, then the discussion form for I j is finished and the process is repeated for the next participant until all discussion forms are produced.
• the discussion forms are received/processed and the discussion tree (13) is updated accordingly.
• the FORM RECEPTION procedure remains active for a specified time, during which the participants are expected to fill up and send back their discussion forms.
• the discussion tree is updated by replacing the empty textual content of the newly created nodes by the content retrieved from the discussion forms.
• the current discussion cycle is terminated and the method attempts to start a new discussion cycle, as described below.
• the NEW CYCLE procedure (11) opens a new discussion cycle.
• the current configuration of the discussion is analyzed and new interactions (assignments) are proposed.
• the analysis of the discussion tree and the proposition of new interactions are the source of intelligence of the method and the principle of the automatic discussion mediation.
• the NEW CYCLE procedure attempts to produce a set of assignments (N, I) that will cause the current discussion tree ( ⁇ n ) to advance to the subsequent cycle ( ⁇ n+1 ).
• FIGURE 3 illustrates the NEW CYCLE procedure and the following paragraphs describe its steps.
• the FILTERING step selects the nodes of the discussion tree that might be worth re-launching, i.e. nodes with non-empty textual content ("answered” nodes) and non-supporting nodes (nodes other than ARG++ nodes).
• the result of the FILTERING procedure is a list of "re-launchable nodes" (20).
• the PARAMETER EXTRACTION step (16) assigns to each re- launchable node a specific "re-launch score", named RS parameter.
• the relaunch score reflects the importance of a node for the discussion and determines the priority of such a node to be re-launched in the subsequent discussion cycle.
• the RS of a node is a real number ranging from 0 (lowest priority) to +1.0 (highest priority).
• the value of RS depends on four sub-parameters, namely: (i) the disagreement level of the node (disagreement sub-parameter), (ii) the depth of the node (depth sub- parameter), (iii) the refuting intention of the node (attack sub-parameter) and (iv) the coverage degree of the node in respect to the group of participants (cover sub-parameter).
• the following paragraphs describe each sub- parameter.
• the disagreement sub-parameter measures the degree of disagreement that the textual content of a certain node exhibits in respect to its direct descendant (child) nodes.
• the disagreement sub-parameter ranges from 0 (no disagreement) to +1.0 (maximum disagreement).
• the disagreement sub-parameter serves to assign higher re-launch priorities to the most polemical nodes, i.e. those with the highest disagreement levels. If a given node N has one or more direct descendant ARG nodes, the disagreement level of N, named disagreement ⁇ ), is calculated as the average disagreement level transmitted by its direct descendants (TS(G)) (see EQUATION 1 and EQUATION 2); if N is a leaf node, the disagreement level of N is assigned 0.
• the depth sub-parameter measures the distance between a node and its corresponding Q-node. It serves to assign higher re-launch priorities to nodes that are closer to the root, i.e. nodes with lower depths.
• the depth sub- parameter of a node is normalized to range from 0 (closest to root) to +1.0 (furthest from root).
• the attack sub-parameter measures de degree of refutation of a node in respect to its parent node. It serves to assign higher re-launch priorities to nodes that exhibit higher refuting intentions.
• attack(N) can assume four possible values: 0 (for ARG++ nodes), +0.33 (for ARG+ nodes), +0.66 (for ARG- nodes) and +1.0 (for ARG — nodes). In the case of Q-nodes and ALT-nodes, the attack sub- parameter is assigned 0.
• the cover sub-parameter measures the degree of group participation in respect to a given node of the discussion tree. It serves to assign higher relaunch priorities to nodes that exhibit low group participation levels.
• the cover sub-parameter is calculated as the ratio between the number of different participants that appear as authors of the direct descendants of a node and the total number of participants of the group.
• RS re-launch score
• the ORDERING step (17) uses the re-launch scores computed in the preceding procedure to sort the re-launchable nodes according to their relaunch priority.
• the result is stored in a list of "sorted re-launchable nodes" (21), where the nodes are distinguished from each other with respect to their relative importance for the discussion.
• the ASSIGNMENT step (18) detects potential peer-to-peer interactions in the existing discussion and proposes new assignments among the participants.
• the ASSIGNMENT step applies specific algorithms, called “assignment mechanisms" (22 ... 26), that analyze the relations between the nodes of the existing discussion ( ⁇ ⁇ ) and propose new assignments for the next discussion cycle ( ⁇ n+1 ).
• the ASSIGNMENT step results in a list of proposed assignments (27) containing the nodes to be re-launched and the corresponding target participants. Each assignment mechanism proposes a specific set of assignments, according to its own interaction objectives.
• EXT externalize
• REPLY BUDDY
• SPREAD SPREAD
• VLD-ATCK VLD-ATCK
• the EXT (externalize) mechanism (22) aims at assuring that all the participants answer all the proposed questions.
• the assignments proposed by EXT are defined as follows. Let Q be the list of questions Q[l] ... Q[m], I be the set of participants I[l] ... I[n] and missing (N) be a function that returns the list of missing participants of N (i.e. the participants who do not appear as authors of any of the direct descendant nodes of N).
• the list of EXT assignments (A ext ) is produced by the following function, assuming that Aext is initially empty.
• the REPLY mechanism (23) aims at assuring the right of response to every participant who had his/her viewpoint refuted in previous discussion cycles.
• the BUDDY mechanism (24) aims at allowing the participants to analyze alternative viewpoints to the issues that they have already reflected on.
• the BUDDY mechanism is defined as follows. For each node N of the discussion tree, the set of direct descendants of N, named chiidren(N) is determined.
• N For each pair of distinct elements of chiidren(N), represented by (N l3 N 2 ), two nodes N' ⁇ and N' 2 are created as direct descendants of Ni and N 2 respectively.
• N' ⁇ is then assigned to the author of N 2 and N' 2 is assigned to the author of Ni. This procedure is repeated for all distinct pairs of chiidren(N) and for all nodes N of the discussion tree.
• the BUDDY assignments are obtained as follows. Let alt-arg-nodes be a function that returns the list of all ALT and ARG nodes of the discussion tree. Let chiid-iist(N) be a function that returns the list of direct child-nodes of the node N. Let buddy(node- ⁇ st) be a function that returns the list of all distinct pairs of elements of the list of nodes node-list.
• buddy-pair : buddy-list [k] ; add (buddy-pair [1] , author (buddy-pair [2] ) to Abuddy; add (buddy-pair [2] , author (buddy-pair [1] ) to Abuddy; endfor endfor return (Abuddy) .
• the SPREAD mechanism (25) aims at assuring that all participants argue over all viewpoints of the discussion. It is defined as follows. For each node N
• ALT or ARG node the set M of all participants who do not appear as authors of any of the direct descendants of N (set of missing participants of N) is determined. For each participant I in M, a textless node N' is created as a direct descendant of N and N' is assigned to I.
• (A sp read) is produced by the following function.
• the VLD-ATCK (validate attack) mechanism (26) aims at assuring that the participants who are assigned the role of tutor of the discussion (if any), validate all refuting argumentations posted by non-tutor participants.
• the assignments proposed by the VLD-ATCK mechanism are defined as follows.
• T be set of tutors, i.e. a subset (possibly empty) of all participants who are assigned the role of tutor.
• G ⁇ either ARG- or ARG —
• Gj' is assigned to the first element of T; this repeats for every remaining element of T.
• VLD-ATCK assignments are defined as follows. Let tutor-list and refuting-nodes be two functions that return respectively the list of tutors and the list of all refuting nodes (ARG- and ARG—) of the discussion.
• the ASSIGNMENT ARBITRATION step (19) receives as input the list of proposed assignments (27), as well as the WL max parameter (28) and outputs the list of final assignments (29).
• the list of final assignments contains only valid assignments that respect the specified maximum workload and assure that the participants have the same number of elements in their discussion forms.
• the ASSIGNMENT ARBITATION step proceeds as follows. Firstly, it eliminates duplicate assignments, which may have been produced by different assignment mechanisms. Secondly, for each proposed assignment (N, I), the discussion tree is searched for the existence of a node N' whose parent is N and whose author is I; if such a node is found, i.e.
• the assignment (N, I) is eliminated from the list of proposed assignments.
• the list of proposed assignments is filtered to eliminate the assignments that exceed the maximum workload per participant (WL max ).
• the list of final assignments is produced. Algorithmically, the ASSIGNMENT ARBITRATION step is described as follows. Let I be the list of participants, let all-nodes be the a function that returns the list of all nodes of the discussion tree and A pr0 p os e be the list of proposed assignments. Let count be the variable that counts the number of assignments of a given participant.
• the list of final assignments referred to as
• a f i na i (initially empty), is produced by the following function.
• the participants should be notified about the opening of a new discussion cycle, for instance, by means of e-mail messages containing customized links to the corresponding discussion forms.
• Such links may trigger a "form building script", on the server side, which dynamically generates the discussion form for the specified participant and cycle.
• The, "form building script” may generate the discussion forms in HTML/XML or similar format, using text areas and radio buttons to allow the participants to type their postings and to express their argumentative intentions (as exemplified in FIGURES 4 and 5).
• the "form building script” corresponds to the DELIVERY step of the NEW CYCLE procedure.
• the NEW CYCLE procedure may be triggered by a scheduler/timer, according to a predefined schedule or a periodic time interval.
• a specific interface should be designed for this purpose. This interface may also be used to set the WL max parameter at each cycle, possibly modifying it to produce variable workloads throughout the discussion.

Priority Applications (1)

Applications Claiming Priority (2)

Publications (2)

Family

ID=39545014

Family Applications (1)

Country Status (3)

Citations (5)

• 2002
  • 2002-04-17 BR BR0201651A patent/BR0201651A/pt not_active Application Discontinuation
• 2003
  • 2003-01-15 AU AU2003201223A patent/AU2003201223A1/en not_active Abandoned
  • 2003-01-15 WO PCT/BR2003/000004 patent/WO2003087970A2/en not_active Application Discontinuation

Patent Citations (5)

Also Published As

Similar Documents

Legal Events