The People’s Bot is a project by Chelsea Barabas and Nathan Mathias from MIT’s Center for Civic Media that uses telepresence robots for the public good.

“Our telepresent bots broaden inclusion at events & support public interest media-making, through scholarships, media fellowships, and auctions.”

And by public good, they mean for example sneaking into purchasing participant registration for The People’s Bot at the highly popular Computer-Human-Interaction Conference and letting people apply for slots. If you’re interested in CHI, but like many people can’t afford to attend, it looks like you can still bid for a much cheaper telepresence opportunity on eBay. And you don’t even have to put on pants. Bid on Ebay to attend the History of Wearables & Google Glass exhibit at CHI (proceeds go to the CHI student travel grants).

Rumor has it The People’s Bot is currently wandering around the Theorizing the Web conference in New York City.

While not everyone here at Robohub agrees that a telepresence device is technically a robot, this is a pretty cool project. We hope with all of our hearts that they don’t get sued.

This post is part of Robohub’s We Robot coverage.

Paper: Risk Management in Commercializing Robots
Author: Stephen Wu
Commentator: Julie Martin

Both Wu and Martin repeatedly made the disclaimer that none of this constitutes “official” legal advice, but it’s pretty darn close. This is a highly practical, specific paper with some helpful insights for robotics manufacturers.

The paper starts with applicable legal theories, covering the range of liabilities that companies could face and the consequences thereof. Wu specifically describes the prominent Ford Pinto and Vioxx cases, and warns against giving rise to so-called “angry jurors,” Cost-benefit analyses made by companies are not always taken kindly by juries, especially when dealing with human lives, leading to very large, indemnifying verdicts.

Wu explains what manufacturers can do to preempt accidents and prevent lawsuits. Martin highlighted the advice on being proactive: going beyond what is required and beyond what is industry practice. Not just sticking to the bare minimum will signal that a company is responsible and concerned with safety. Wu also draws attention to some non-traditional risks, saying that [1] information security threats (hacking, data privacy) and [2] supply chain risks (financial fraud, security, reliability) may not be sufficiently taken into account by robotics manufacturers. Further advice in the paper includes documentation, risk management technology, helpful standards and organizations, etc.

In the discussion, Martin specifically brought up the problem of data security, stressing the importance of privacy to consumers. Technologies like telepresence robots, with recorded data, facial recognition, location tracking, etc. raise questions of privacy by design. She suggested that the framework of the paper may also be a good framework for liabilities arising from privacy issues.

There was some discussion of what jurors tend to take into account (facts, reputation, etc.), as well as the question of disclaimers and their limitations. The audience also inquired what was robotics-specific about these liability risks and ways of managing them. Martin and Wu both agreed that the general questions are not necessarily unique to robotics. But Martin made an interesting point that people might have a more visceral reaction to these issues when applied to robots. She said that this may even raise awareness and incentivize desirable risk management behavior that people should be thinking about in other industries but currently are not.

Even if automated cars are twice as good as human drivers, someone asked whether it could be a problem that companies are releasing technology that they know will more directly be responsible for a certain number of lost lives. Wu replied that manufacturers will emphasize not the number of deaths they’re taking into account, but rather the number of lives they’re saving through the technology. Martin added that many other industries have thrived in the face of similar risk management (e.g. pharmaceuticals).

Someone also brought up the development of robots that foster affection in humans and asked whether this might someday give rise to a separate liability issue. E.g. A lawsuit based on addiction to a robot’s affection and the emotional trauma of having that removed. While the audience was audibly amused, Wu agreed that such future scenarios are likely.

See all the We Robot coverage on Robohub →

See also:

• Will lawsuits kill the autonomous car? | MSN
• Robo-cars face a new threat: Lawyers | CNET
• Robot wars: With the technology in place, inventors examine legal hurdles | Nature World News

This post is part of Robohub’s We Robot coverage.

To everyone’s delight, the first day wrapped up with a presentation by Patric Verrone, co-producer of the popular TV-show Futurama.

Presentation: Bender’s Law
Speaker: Futurama Co-Producer Patric Verrone

Anyone familiar with the show will know that it is a goldmine of commentary on politics and robots. Verrone, who has a legal background, spoke about the ways and various instances the show thematizes robots and the law. He is funny as hell and showed some great Futurama clips, which unfortunately are not visible in the video above. Presumably for legal reasons. Which generally speaking is sad, because how could this possibly be anything other than good promotion for Futurama? In any case, blame the system, not the players, and watch the show if you haven’t! It’s awesome.

See all the We Robot coverage on Robohub →

The program committee of the inaugural Drones & Aerial Robotics Conference (DARC) is now inviting proposals from individuals across a broad spectrum of expertise. DARC will convene a wide range of stakeholders to explore civilian drones and their impact on society. The conference is presented by the Engelberg Center on Innovation Law & Policy at NYU School of Law and will be held October 11-13, 2013. We invite contributions from all interested disciplines on or before May 27, 2013.

ABOUT DARC

What do hobbyists, journalists, university researchers, soldiers, and police all have in common? They’re all testing or using aerial robotics, commonly called “drones.”

Imagine a near future in which networks of autonomous robots roam the skies, performing everything from law enforcement, to communications, to crop dusting, shipping and logistics. Sound implausible? It might be. But that’s the future that the aerospace industry, the FAA, and a new class of entrepreneurs are busy preparing. And this future is closer to reality than many realize. Drones force a broad reconsideration of the laws and regulatory frameworks that protect vital interests like civil liberties, due process, privacy, innovation, and security. But they also herald new innovations and new public interest applications.

These technologies are powered by some of the same phenomena that propel the mobile industry—Moore’s law, economies of scale, and ever more sophisticated software built by large companies or open source hackers. Anyone can buy a toy drone at the mall for $299, assemble a relatively capable DIY drone for $1000, or buy a sophisticated $50,000 aircraft for amateur photogrammetry.

As the FAA scrambles to meet its 2015 integration deadline—which will open domestic airspace to hobby aircraft as well as the types of aircraft that currently fly over Afghanistan—DARC provides a forum for entrepreneurs, policy makers, and civil libertarians to constructively engage. Join us this Fall to explore this fascinating, emergent space.

PROPOSALS

There are few technologies that stimulate such deep and wide-ranging questions about law, technology, and international relations. We welcome all proposals, but please stay substantive and constructive. This list is not exhaustive, but we invite proposals along the following lines:

• Anecdotal talks
• Popular research
• Demos or tutorials
• Scholarly research
• Constructive discussions

Proposals should address — but are not limited to — our five main areas of interest:

• General Interest & Zeitgeist
• Makers & Entrepreneurs
• Law & Regulation
• Privacy & Surveillance
• Security

Proposals can be submitted via the conference website at https://droneconference.org/proposals. Proposals will be accepted until May 27, 2013. Questions may be sent to contact@droneconference.org

This post is part of Robohub’s We Robot coverage.

Robot Demo
Speaker: Ian Danforth

What was billed as a robot demonstation was coupled with a talk by Ian Danforth on household robots.

After introducing the audience to his robotic companion, Danforth spoke about a current rise of personal robots. What we had been waiting for previously, he says, were factors like science (machine learning, neuroscience), computation, sufficient bandwidth, affordable components, a global market, societal readiness (i.e. for people to get over the assumption that all robots look like Rosie), and sufficient data (because robots require lots of data to capture human-like experiences). But now that we’re seeing all of the above come into existence, he estimates that there will be an explosion of home robotics within the next three years.

It starts small
It starts cute
It starts now

Why small and cute? Danforth says because of expectations. Designing household robots to be adorable can create a tolerance where the expectation would otherwise be that the robot perform flawlessly. We never expect our pets to act perfectly, he says, or small children. And when they don’t, sometimes we actually like it. It’s an intrinsically enjoyable experience to try to teach something that is cute and small and makes mistakes. The “now” part stresses that when we talk about home robots, we are talking about existing technology. “This is not the future – this is today.”

Five years from now, should we want our children to have robotic pets, or Lassie? Danforth says it depends. “Is your kid allergic to Lassie? Do you mind cleaning up poop?” In the United States, 4 million cats and dogs are killed every year. For the percentage of people who would like to be pet owners, but can’t be or shouldn’t be, artificial pets may be a viable alternative.

Someone asked whether we should give these robots “rights”. Danforth asserted that people will certainly want to. This could start on the level of company policies, e.g. if you abuse your pet, that could be something that causes you to lose access to the service that enables it. But he thinks that it will be a long time before the law recognizes any artificial entity as sufficiently complex to deserve legal protection.

When asked about data collection and privacy, Danforth said that he as a developer is aware of the large amounts of different data he is collecting, and that he is thinking about how to ensure people’s privacy (e.g. by encrypting video and audio streams). The two challenges to address: 1. How informed is the end-user about what happens with what data? 2. The technological responsibility of preventing unnecessary data retention and developing sufficient security around what’s necessary. But he also expressed the hope that non-developers were thinking more deeply about these issues and would be able to help him both “be a radical innovator” and “not get sued.”

The discussion also covered expectation management, projection, artificial intelligence, character gender, and different methods of encoding personality.

At some point during Danforth’s talk, his robot somewhat impolitely fell asleep. The rest of the audience most certainly did not.

See all the We Robot coverage on Robohub →

This post is part of Robohub’s We Robot coverage.

Woodrow Hartzog and Greg Conti presented their paper (co-authored with Lisa Shay and John Nelson of West Point) on implementing law in computer systems.

Paper: Do Robots Dream of Electric Laws? An Experiment in Law as Algorithm
Author: Greg Conti, Woodrow Hartzog
Moderator: Harry Surden

This was an especially interesting presentation, given that last year at We Robot in Miami, the same authors presented a paper on the problems of removing humans from the loop and turning law enforcement over to a computer (Confronting Automated Law Enforcement). This year, the authors took the question a step further. In an experiment, they look at what happens when you convert laws into algorithms. It turns out that even simple rules (like speeding limits) require unexpected consideration.

In the experiment, 52 programers were asked to automate the enforcement of traffic speed limits. They were given a set of driving data and each wrote a program that measured the number of speed limit violations and issued traffic tickets accordingly. Despite having exact data for both vehicle speed and speed limits, the number of issued tickets varied among the programs. The authors attribute the variance to the fact that the programers were faced with having to make assumptions and legal interpretations, for example whether to code according to the letter of the law (100% enforcement of every violation) or the intent of the law (tolerating minor infractions). The study indicates that there can be unanticipated degrees of freedom in the design of enforcement algorithms, even when dealing with seemingly straightforward legal rules.

As for the broader implications, the authors drew attention to the following problem areas:
Culpability We may not be able to automate laws that include culpability, because this is too difficult to determine automatically.
Objectivity of the wrongful conduct E.g. speed does not necessarily equal recklessness.
Identification With automated enforcement we would have to make particularly sure that someone’s identity can be verified.
Accessibility To what extent does this lead to a surveillance society? What kind of information is and should be accessible?

The results of the study also caution against outsourcing coding of the law to third-parties. They indicate that seemingly minor details can make a huge difference, creating more variance than assumed and requiring decisions by someone with actual rule-making authority. The authors also mentioned the need to consider potential second and third order effects, such as on traffic flow, etc.

Surden pointed out that there are already examples of laws being translated into computer code today that many people, including law-makers, are not fully aware of. For example, tax preparation software Turbotax inherently involves a lot of judgment and design decisions that have gone under the radar and are accepted more or less unquestioned by the Internal Revenue Service.

To my (personal, IP-nerdy) delight, an audience member mentioned the problems with content ID-ing and automatic copyright takedown notices on platforms like YouTube. Often these programs will remove completely legitimate content, causing troubles for fair-use activity. Surden said that one take-away for automated enforcement from this is that the appeal process should be just as easy as the take-down process.

The discussion also turned to the question of socially desirable enforcement. If we can embed laws in systems, we could potentially have a system of perfect monitoring and perfect enforcement. But are there costs to that perfection? Both authors argued that imperfect enforcement allows for a healthy amount of discretion and flexibility, and that bureaucracy left to its own devices could be dangerous. Even more problematic than perfect enforcement, said Hartzog, is perfect prevention. As we’ve seen with red light camera systems, citizens can become outraged by what they perceive as a lack of value judgment. A disconnect between law and social perception can go so far as to be counterproductive. Conti also postulated that the type of person who is willing to take on some risk to achieve a certain gain would be beat out by perfect law enforcement – behavior that we might want to otherwise maintain in our society.

The audience was also interested in the question whether our current system of human error, bias, and corruption in law enforcement is better or worse than a system that leaves no flexibility for human judgment. Can we build room for discretion and community norms into systems? What happens if a human bias becomes systematized? Interestingly, when the programmers were asked after the experiment whether they would want to drive on the roads with the programs they had built, their answer was uniformly “no.” (Although one programer said yes, conditioned on creating a backdoor exception for herself.)

In terms of policy recommendations and reducing the uncertainty of coding law, the authors suggested creating a committee or organization to set standards for automated enforcement systems, and also that the code of such systems be kept transparent and open to examination.

See all the We Robot coverage on Robohub →

This post is part of Robohub’s We Robot coverage.

The conference kicked off with a panel on intellectual property (IP) in the robotics industry.

Panel: Robotics and Intellectual Property
Speakers: Mark Lemley, Christian Hicks, Julie Samuels
Moderator: Brian Biddinger

The panel discussed general intellectual property issues (in particular patents and copyright) and how these could apply to emerging technologies in robotics. If you’re interested in learning a bit about the current problems with the patent and copyright system in general, this panel is worth a watch.

Regarding the robotics industry specifically, the panelists distinguished between hardware and software, and general purpose technology versus special purpose technology. They also discussed how robotic technologies may start out specialized and proprietary, but become increasingly commoditized over time – something seen previously in the computer and smartphone industries. The panelists’ advice on IP strategy (for individual companies as well as the industry in general) largely depended on the types of technology being developed.

For example, whether a company is devoting their main resources to hardware or software will determine how much of their IP life is influenced by patent or copyright law. The panel came down hard on software patents, arguing that they have done more harm than good and appealing to the robotics community to do everything possible to prevent them. Hardware patents were more difficult to dismiss, despite the danger of creating an unnecessary and suffocating patent thicket. As Samuels commented “I wish I could say ‘ignore patents!’ But they won’t ignore you.”

The panel speculated that specific purpose technology may lend itself to proprietary models, but robots that can later on be programmed by other people to do a variety of different things might be better to release as open-source. Comparing to other industries, Lemley observed that the companies that start out closed and proprietary are often first-movers, but they also often end up an edge player in the market later on. An opposite example is the Arduino model – releasing your platform as open, and focusing instead on branding. Because it’s hard to anticipate all of the things that people will want to do with the platform, it can be a successful strategy to keep it open at an early stage – so long as the brand is strong enough. (E.g. manufacturing robots and surgical arms may not necessarily be at that “early stage,” but perhaps household robots that will be able to do a variety of yet unknown things.) Hicks pointed out that interoperability may become increasingly important as robots start working together. Having a closed-off system like Apple can only work if you’re extremely and unusually successful – in any other case the better strategy for such an emerging eco-system will be open.

Hicks also spoke about his paper Robots, the DMCA, and Patents: Threats, Strategy, and Case Law in the Aftermarket, which describes how hardware manufacturers might start using the anti-circumvention provisions of the Digital Millennium Copyright Act (DMCA) to control secondary markets for service, software upgrades, etc. Based on the precedent of two court decisions, the DMCA can potentially be (ab)used to prevent customers and competitors from accessing software, and thus maintaining, upgrading, and adding functionality to robots, or even analyzing a product for patent infringement.

The panel talked about the problem of “trolls,” pointing out that over 60% of all lawsuits last year were brought by non-practicing entities (85% of these based on software patents). The panelists discussed (and disagreed about) the importance of patented technology for startups in getting VC funding. They also debated longer-term patent and copyright reform. As for the immediate future, the panel recommended that the industry 1. stay informed and share information with each other; 2. get involved and look for solutions within the current system, such as open-source software, getting behind projects like the defensive patent license, or working with agreements like Twitter’s Innovators Patent Agreement.

Other mentioned links of interest:
Lex Machina
Electronic Frontier Foundation

See all the We Robot coverage on Robohub →