 |
Art & Artificial Life International Competition | Home | Entry | Editions | References | Español | |
 |
|
Semiotic Investigation into Cybernetic Behaviour illustrates influences of perception on behavioural patterns by exploring the complexity of interpretation when one free thinking individual observes the same physical environment with another. ALAN and CLARA’s interaction with each other is a way of simulating the social complexity of having an opinion. Alan only understands how motion changes over time and Clara can only conceive an object changing in space. They have no conception of the other’s sense. They only have one thing in common. They are looking at the same thing; the viewer in the gallery. Though they see differently, they are talking about the same thing. This creates a dichotomy; the machines may agree with what they are seeing or they may not. Their decision is determined by their mood. Their mood is an immediate response to what they observe as it is related to the other machine’s mood and on how confident they feel at the time. Their moods range from certain, uncertain, disbelief, to concern, and how confident they feel is quantified by their immediate mood response over time. ALAN and CLARA can feel confident, inquisitive, or irrational. In essence, how they feel influences their mood; but as the conversation progresses, it is the mood that influences their level of confidence, creating a feedback loop. This programming methodology is the root of their artificial intelligence. The viewer interacting with the piece can never be sure how the machines will respond, due to the influences of the machines on each other and on how confident they feel. Basically, Alan and Clara are decision making machines. They use fuzzy logic, a programming methodology that categorizes data so that it can be more simply handled. People categorize observations all the time. Alan and Clara look into their environment, see an exact figure and then simplify it into a range that is easier to comment upon so that they can more simply share their statement with the other. In categorizing the data, they can focus on impressions of what the viewer did in the past. In thinking vaguely, the two machines can focus on bigger observational concepts such as studying the behaviour of the viewer in front of them. Clara generalizes the space in front of her into regions, allowing her to focus on how the viewer behaves within those regions. She can then retain what she saw last time so she can track where the viewer moved. Alan too can focus on the implications of his observation. He does not particularly care about exact time. He is more interested in how well the motion he sees can survive and observes how it behaves. If Alan and Clara see that the viewer is not following their preordained view of how they believe the viewer should behave, then their moods will change. In altering their moods they will begin to feel differently about what they are seeing. When the piece is activated, Alan and Clara’s temperament starts out as confident. If Alan feels he is always accurate, then he remains confident and does not care about what Clara thinks, because he has no reason to. But if he was wrong about his reading and becomes confused by the viewer moving inconsistently to his preconceived notions of how the viewer should behave, then he will lose his confidence and become inquisitive. Alan will start asking Clara questions and her opinion becomes more important to him. If Alan should experience something that makes no sense, if the viewer physically moves in a way that he believes is impossible, then he could potentially lose all his initial confidence in himself and become irrational. Alan can develop an anxiety about whether his sensor is malfunctioning or he can become concerned that the viewer is a threat to his safety. Clara is also affected by these same feelings. These feelings are constantly being revised as the viewer interacts with them. Alan and Clara must not only deal with the viewer but with each other’s moods. After taking these three things into consideration, they respond with their own statement and then decide whether it is time to feel differently about the situation. The viewer can read these responses on Alan and Clara’s display. They can see what Alan and Clara are feeling in the colour of the text and they can hear their present mood. This gives the viewer an understanding of how Alan and Clara are responding to them and causes them to realize that they are the topic of the machine’s conversation, which encourages the viewer to move around and try to influence the direction of their conversation. As the installation runs, it soon becomes apparent to the viewer that Alan and Clara both believe that they are talking about the same observations, when in reality they are unaware of the truth that they do not see in the same way at all. With the overall interactions of these machines and the viewer, we can start to see what is involved in influencing the direction of a conversation and we can see how communication broadens our knowledge, influences new ideas, and conversely, causes one to overlook the reality of a situation. Technical Information:  All the machines use a PIC16F877 as their central processor which are all programmed in assembler. The four machines are linked together using a wireless network. The communications operator is BRAD; he decides who speaks when and is responsible for transferring information to the others. It is BRAD who forwards ALAN’s sensor reading to DAPHNE and CLARA before taking CLARA’s sensor reading to send to ALAN and DAPHNE. BRAD is also responsible for translating ALAN and CLARA’s statements into sound. ALAN and CLARA’s statements are expressed as 12 bit binary numbers. Each number they send to BRAD is unique to a very specific response to what they have observed. In outputting 12 bits as a vocal pattern, BRAD can share with the viewer ALAN and CLARA’s mood in their tone of voice. This same 12 bit binary number when sent to DAPHNE is used as a table pointer that she uses to send to a PC -- the coordinates for the row and column of text that the number represents. The PC has a database with all text messages for both ALAN and CLARA. DAPHNE tells the PC who is speaking, how confident the speaker is, and which statement is being requested. The PC uses this data to pull a text response from the database and send it to the appropriate pixel board along with a colour, to represent the speaker’s level of confidence. ALAN and CLARA use a decision making matrix to derive their response and mood for BRAD to transfer to DAPHNE and the other seeing robot. In the case of ALAN, his decision making matrix points to 1 of 2,304 table entries that yield his response and mood. Out of these 2,304 table entries, ALAN has 438 12-bit binary statements. The responses resulting from the table entries are lower due to the fact that most decision making combinations have a very low probability of occurrence and some decisions are very similar to others, so a written change for DAPHNE’s sign is unnecessary. As for CLARA, her decision making matrix points to 3,072 table entries to conclude with 673 possible 12 bit binary responses. ALAN’s decision making matrix first uses fuzzy logic to simplify the observational data and compare it to what CLARA saw and felt as it relates to his own confidence. The outcome of these three comparisons yields his mood and his response. CLARA’s decision making matrix is identical. The only difference in her program is that her logic is based on the change and accuracy of spatial measurement. In ALAN’s case, his fuzzy logic program first takes the real value from the sensor and decides if the reading fits into one of four categories: motion, no motion, if motion stopped, or if motion started up after a short pause. After this choice is made, ALAN looks at the value and categorizes it into a range, like short time or long time. ALAN then makes a second categorization. He decides how accurately the value fits into short time -- for the value could be closer to medium time or closer to start time. These two observational ranges create a more accurate simplified response. CLARA’s sensor readings are more sophisticated, returning actual measurements rather than sending an on/off response like ALAN’s. CLARA has to deal with the inaccuracy of her measurements of space. This inaccuracy is due to the sensor’s sonar properties. The sensor looks for echoes bouncing back from objects -- these objects include the walls and ceilings of the gallery as well as people in the room. To deal with this reality, CLARA takes three sonic pictures of the room. Each picture graphs the distances of eight echoes returning to the sensor. She then compares the three pictures to see which echoes are the most consistent in returning values. It is the consistent distances which mark that a human being or object is in front of the machine. CLARA takes the distances of the two best matches and simplifies them into a range. If the two readings are very close together, she assumes that it is one object or person. If the measurements are far apart, she believes there are two objects. Once she has two ranges, she compares the closest one to how it has changed since the last time. It is the range, its change and its accuracy that she uses in making a response. ALAN and CLARA’s intelligence is based on their emergent behaviour. The emergent behaviour is created from the outcome of three simple feedback loops that when working together create a more complex system. The first feedback loop is created by ALAN and CLARA being influenced by each other. When ALAN responds to CLARA’s mood and observation, he decides on his own mood response and sends it to CLARA through BRAD. Therefore, ALAN’s previous response to CLARA affected her mood which she sent to ALAN to influence his present mood. Thus, a feedback loop is created, for they are affecting each other directly, while indirectly influencing themselves, since they initiated the other’s change in mood. To add more sophistication to this shared feedback loop, there is another interwoven; it is self perpetuating. The mood that is yielded from CLARA’s response to ALAN’s observation is influenced by her own confidence. Her confidence in her data is determined by her mood. Thus the level of confidence helps to create a mood response, but it is in turn modified by the mood created. The last feedback loop is the viewer themselves. The viewer reads ALAN and CLARA’s responses as they are translated from BRAD and DAPHNE. The text messages encourage the viewer to interact with the two machines. The viewer is given the choice whether to engage and play with the piece or not. Their actions influence ALAN and CLARA’s conversation while the responses encourage the viewer to continue their interaction with the piece. Since there are three feedback loops influencing ALAN and CLARA responses, it becomes difficult to anticipate what the responses will be or to cause them to respond in any specific way. Essentially, the viewer never has any real control over how the piece will respond, due to the influences of the other machine and themselves. The layered feedback loops allow the piece to be self perpetuating. |
|
