A new desk for our new President
It is time to build a 21st century Presidential desk that can run on any internet-connected computer or television. A new Presidential desk would provide the President and every other citizen with a slick new user interface to the entire federal government, and it would help with the White House correspondence problem.
In the 2008 campaign, I served as the Targeting and Modeling lead (the staff statistician) at the Obama campaign headquarters. After we won, I floated the idea of the President’s desk. It gained support from our tech-heavy staff, until somebody with political sense pointed out that launching a national open source effort to build our new President a new desk would be a pretty tone-deaf thing to do with our economy in free-fall. So now, with our country in or near to a constitutional crisis 114 days into the next President’s administration, I feel we’ve arrived once again at a similarly perfect moment to discuss building a new desk for our President.
The marketing conceit of the Presidential desk could be that the President should be able to log in any time anywhere and make Presidential orders. The image of President Bush reading to 3rd graders when he learns that the twin towers have been attacked can be associated with a replacement fantasy in which he goes over to the classroom Apple II and orders our Air Force to scramble. In truth, the President always has great mobile technology available, and there is never a need to use just any computer at hand, but the idea of any computer being the one that sends the orders may spark the imagination.
More importantly, a White House interface to the Federal government would provide helpful organization and a common vocabulary for inhuman volumes of information.
There are already pieces of the puzzle in place. A central interface will help validate and coordinate those pieces. The Office of Management and Budget (OMB) provides a standardized, annual cross-government program evaluation using a standardized tool (the Program Assessment Rating Tool). Every agency provides routine reports to Congress and annual budget submissions. Program evaluations and budgets, agency org charts and websites, all form a kind of de facto structure. But I suspect nobody has read and digested all this information…. nor kept up with the daily changes.
Information about our government needs to be organized to be understood, but there does not appear to be any structure underlying or explaining the ad hoc organization of the agencies. The House and Senate each have somewhat different committee structures to oversee the work of the agencies. For example, the House has a committee for Energy and Commerce; the Senate has a committee for Energy and Natural Resources and another for Commerce, Science, and Transportation. The House and Senate committee structures help us attach meaning to the “60 to 316 federal agencies”, but nobody seems to know how many there are, much less how to organize and understand what they all do. The lack of cross-government coordination and central structure became clear after 9/11 when we saw that there were many distinct organizations tracking potential terrorists. The response was to create a central office, the Director of National Intelligence. This new bureaucratic layer did not solve the underlying problem for intelligence community, much less the broader problem for the government.
The Presidential desk will provide structure and semantic (by-meaning as opposed to by-name) indexes into the disparate government activities. For example, assistance programs can be found at benefits.gov. This improves on by-name indexing in which it was only possible to find benefits programs if you already knew the names, like SNAP , CHIP, and TANF. The Presidential desk would be a central coordination point by which any federal program could be found by way of its purpose. The effort to build the Presidential desk would assist in the overarching need to create a central semantic index into all those programs.
While I am quite serious that a new Presidential desk would be helpful for the White House in understanding what is going on, I am most interested in the Presidential desk for what it will do for private citizens. To be clear, a modernized desk would not actually control the government. There is no need to supplant the existing channels of executive orders. The problem flows in the other direction: neither the President nor any of us really knows what the agencies are or should be doing. There’s no common vocabulary and (therefore) no common tracking mechanism. I do not care about tracking and controlling troop movements and spies. I care that any of us can use the desk as an interface to the unclassified preponderance of our government.
So, let’s sit down as private citizens at our brand new Presidential desk and take it for a spin. Suppose we want to add a high-speed rail line from San Francisco to Los Angeles. The interface has a US map. We click on the map to expand it and several buttons show up at the bottom. One of the buttons gives us viewing options so we can see super-positions of pollution levels, transportation lines, tax income, what have you. There are so many options, we may need to use the voice/text search interface to help us find what we want. Similarly, another button gives us an enormous range of build options. We can build a border wall, re-zone national parks, add a new face to Mt. Rushmore, anything our monstrous little hearts desire. Today we use the build menu to draw in a new high-speed rail line.
Once we’ve drawn in the rail line, we can use another button to launch policy simulations and analysis. We can see the projected cost and timeline for the build and how the project will impact the local economies. We can simulate the build and operation of the rail line; we can control the time scale, spatial scale, outcomes being simulated, and the assumptions surrounding the simulation. We can share the simulation and play it as a game and/or study the simulated outcomes in detail to establish policy. The desk is fun! The desk is useful!
And the desk motivates a cross-government data architecture. We had the button to build the high-speed rail line because a while ago somebody wanted to add a high-speed rail line and noticed there was no button for that. They only had to name it and voila, the map had a new (empty stub) button. Over the next few months several more people wanted to use that button and noticed it needed some work. A volunteer contributed a graphic for the rail lines; other users added outcome variables they wanted to simulate; and, due to the level of interest, responsibility and budget for the high-speed rail feature was assigned to the Department of Transportation. This assignment automatically triggered several tasks that individuals or businesses could perform for a small reward to fill in the simulation logic. The first developers were paid, no questions asked. As the feature became popular, the contracts to improve simulations became bigger and more competitive. Over time, the high-speed rail feature grew to include more actions, projections, advanced simulations, and academic studies.
In addition to evolving a cross-government data architecture, the President’s new desk will substantially help the White House Office of Presidential Correspondence, and other agency’s correspondence offices, with the correspondence problem:
1) Translation. The first problem in constituent correspondence is making sense of what the constituent is saying. The only person who really knows what the citizen is trying to say is the citizen him/herself. It would be nice sometimes to go back and ask, “what did you mean here”? The desk provides a fun and useful way to do that.
The desk converts citizen interests and intent into specific actions that are the purview of specific agencies and agency staff. A citizen interacting with the desk will have the option of providing a message, like a proposed speech to the nation, as well as actions that illustrate the point. A link to the desk, or perhaps subsets of actions within the desk, can also be sent in response to speech/text/video messages sent by the citizen. Whether the desk is receiving primary search text or actions meant to explain some initial correspondence, the desk serves as a translation device. The citizen’s verbally stated or written intent is associated with the citizen’s functional goals. The desk will provide an invaluable training/testing corpus for machine translation.
By associating citizen communications with representations of meaning that are expressed in terms of what the government can actually do, the desk will enable machine learning to construct systems that interpret messages to the White House and/or agencies with better functional precision than is possible with the keyword-based approaches currently on the market and better than the text mining approaches the market will soon provide. When a citizen writes or says what they want, and then clicks on what they perceive to be the correct desk actions to help them do what they want, we gain new insights into how Americans think and talk about government actions. Automated ontologies will not have that “this is what I meant” information.
2) Categorization. A secondary problem in constituent correspondence is providing the representative with a report on what the constituents are saying. This is usually done by categorizing and counting. The Presidential desk puts much of the work of categorizing the constituent’s communication on the constituents. My previous communication would be categorized as something about high-speed rail, because that’s what I ended up doing.
The purpose of categorizing constituent communications is to help the political leadership track the interests of constituents. It is appropriate for internal reports to be guided by political needs and decisions — the political staff can say whatever they want to each other. However, it is not appropriate for the vocabulary of what is possible to belong to the political staff. If a new President absolutely does not want to build a high-speed rail line s/he can ignore any communications requesting a rail line, but the President should not control our ability to make the request. A President could eliminate the environmental protection agency, but that would make our ability to request water quality metrics and clean water protections even more important.
I conclude that President’s desk must be free and open source. We the people must be able to access and extend the code. The text and speech corpus of Americans speaking to their President must be controlled and protected by the government, but the code snippets that empower the President’s desk must be open. I see the President’s desk as a platform that supports coordination by citizens to advocate for government action. The President’s desk should be integrated with equivalent systems for state houses and Congressional offices. For control, for coordination, for integration, the system should be free and open source.
3) Response. Another secondary problem in constituent correspondence is providing the constituent with a response. The Presidential desk helps with response by bringing the citizen into the process of tracking what is actually occurring with respect to his/her interests. Some administrations will be more transparent than others, but all will want to display what they are doing at times. As tools for creating interactive VR and other advanced content types improve, the government agencies’ public relations will also improve in their ability to depict the standard operations and aspirational goals of the programs they manage and radically transform the process of comment on proposed regulations [coming soon].
With or without the Presidential desk, now-standard artificial intelligence can provide helpful responses to many standard requests. For example, a student may use the desk to open the President’s calendar to invite the President to come speak at graduation or to request a birthday card for her great grandfather’s 100th birthday. These requests can be readily parsed and would lead to appropriate responses. Tweets, letters, phone calls, and videos with these requests can also be processed automatically.
4) Re-conceptualization. A tertiary problem in constituent correspondence is to figure out when to do something that was not previously planned in response to a citizen request. The Presidential desk could include top-level folders for big initiatives such as speeches the President should give. With respect to actual actions the President might take, and all else being equal, these modules would have as much or as little influence as the current White House website’s petition module (https://petitions.whitehouse.gov/).
AI can help the White House (and other offices) recognize how ideas are spreading. Recommendations in letters, petitions, videos et al can be mined automatically for common themes. For example, the current top petition at whitehouse.gov calls for President Trump to release his tax returns. These themes can be cross-referenced to calls, emails and other communications. The White House can compute the demographic distributions (and possibly psychographic distributions) of the petition signatories, callers, and email address owners. If these computed distributions show that the calls for the President’s tax returns are spreading beyond the usual suspects, the White House may want to take action.
The normal operation of the President’s desk will help give structure and meaning to outside-the-box ideas in ways AI cannot. The desk helps us as citizens to collaborate to define our future — for example by creating simulations of the economic impact of a new high-speed rail line. AI can mine text for themes, and it can find semantic grounding in relationships to existing legislation, but AI cannot create meaning for us. The President’s desk will be a collaborative arena in which we can envision new futures together, and that is an essentially human function.
Given present circumstances, this is probably the worst possible time to launch a collaborative, open source effort to build a new desk for our President. Let’s do this!
If you liked this article, please click on the heart below.